application letter image

Application Letter royalty-free images

263,474 application letter stock photos, vectors, and illustrations are available royalty-free for download..

Our company

Press/Media

Investor relations

Shutterstock Blog

Popular searches

Stock Photos and Videos

Stock photos

Stock videos

Stock vectors

Editorial images

Featured photo collections

Sell your content

Affiliate/Reseller

International reseller

Live assignments

Rights and clearance

Website Terms of Use

Terms of Service

Privacy policy

Modern Slavery Statement

Cookie Preferences

Shutterstock.AI

AI style types

Shutterstock mobile app

Android app

© 2003-2024 Shutterstock, Inc.

• AI Generator

job interview

2,870 job application letter stock photos and high-res pictures, browse 2,870 job application letter photos and images available, or search for job interview to find more great photos and pictures..

We’re sorry, but Freepik doesn’t work properly without JavaScript enabled. FAQ Contact

• Notifications
• Go back Remove
• No notifications to show yet You’ll see useful information here soon. Stay tuned!
• Downloads 0/60 What is this?
• My collections
• My subscription

Find out what’s new on Freepik and get notified about the latest content updates and feature releases.

Job Application Letter Images

• Add to collection
• Save to Pinterest

• table flatlay

• business process
• organization

• desk top view
• desk flatlay

• job application
• career jobs
• job interview

• property agent
• estate agent
• property management

• office paper
• office documents

• couple ring

• psychological health

• curriculum vitae
• cv template

• contact information
• cover letter

• digital business
• business connection

• letter template

• money template

• business contract

• children education

• psychologist

• table top view

• human resources

• lawyer office
• legal document

• business data
• business analytics
• business marketing

• paper table
• business desk

• recruitment agency

• offer letter

• sign contract

• formal letter

• education design

• resignation

• work illustration

• computer keyboard
• blank notebook

• car template

• PRO Courses Guides New Tech Help Pro Expert Videos About wikiHow Pro Upgrade Sign In
• EDIT Edit this Article
• EXPLORE Tech Help Pro About Us Random Article Quizzes Request a New Article Community Dashboard This Or That Game Popular Categories Arts and Entertainment Artwork Books Movies Computers and Electronics Computers Phone Skills Technology Hacks Health Men's Health Mental Health Women's Health Relationships Dating Love Relationship Issues Hobbies and Crafts Crafts Drawing Games Education & Communication Communication Skills Personal Development Studying Personal Care and Style Fashion Hair Care Personal Hygiene Youth Personal Care School Stuff Dating All Categories Arts and Entertainment Finance and Business Home and Garden Relationship Quizzes Cars & Other Vehicles Food and Entertaining Personal Care and Style Sports and Fitness Computers and Electronics Health Pets and Animals Travel Education & Communication Hobbies and Crafts Philosophy and Religion Work World Family Life Holidays and Traditions Relationships Youth
• Browse Articles
• Learn Something New
• Quizzes Hot
• This Or That Game
• Train Your Brain
• Explore More
• Support wikiHow
• About wikiHow
• Log in / Sign up
• Job Application Documents
• Cover Letters

How to Write an Application Letter

Last Updated: May 6, 2024 Fact Checked

This article was co-authored by Alexander Ruiz, M.Ed. . Alexander Ruiz is an Educational Consultant and the Educational Director of Link Educational Institute, a tutoring business based in Claremont, California that provides customizable educational plans, subject and test prep tutoring, and college application consulting. With over a decade and a half of experience in the education industry, Alexander coaches students to increase their self-awareness and emotional intelligence while achieving skills and the goal of achieving skills and higher education. He holds a BA in Psychology from Florida International University and an MA in Education from Georgia Southern University. There are 7 references cited in this article, which can be found at the bottom of the page. This article has been fact-checked, ensuring the accuracy of any cited facts and confirming the authority of its sources. This article has been viewed 169,801 times.

Application letters are typically written to accompany school or job applications. The purpose of the letter is to introduce yourself to the decision committee, and to outline your qualifications in a specific way. It can be the only time other than an interview that you have a chance to really stand out in an application, so it's important to get it right. You can learn what to include in your letter, how to style it, and how to format it to give yourself the best chance.

Application Letter Templates

Writing a Job Application

• A good example would be: "I'm writing to apply for the Chimney Sweep position advertised in Rolling Stone. I think my experience in the heating industry makes me uniquely qualified for this position. Please find my application materials and a brief description of my qualifications below."
• Don't write your name until the signature. It'll be in the header and in the sign-off, so there's no reason to put it in the body of the letter itself.

• Be specific. Who are you? Where do you come from? What's your story? These details are important. HR screeners read hundreds of these.
• Describe your ambitions. Where do you want to go? How will this opportunity help you get there?
• What skills and experiences make you the right fit? Be as specific as possible and avoid vague language. It's better to describe a time you solved a specific problem at your last job than to just write, "I'm a good problem solver at work."

• Tailor it to the business. If you're applying to work at a record store, you need to talk about music. If you're applying to work at a tech company that writes, "Tell us something totally rad about yourself!" it's probably ok to be a little more informal.

• Don't over-promise. Telling someone that you can guarantee that you'll be able to turn around their sales figures in six months or less is a good way to get fired in six months.

• Any kind of job requires this type of research. If you apply to a restaurant, you need to be familiar with the menu and the kind of customers the restaurant attracts. Consider eating there a few times before you apply.
• Don't show you're familiar by criticizing a business and telling them what you can do better. Not the time to offer a harsh criticism of a business plan that you don't really know anything about.

Writing a School Application

• Common prompts include things like, "Outline your qualifications for this position" or "In writing, explain how this position would affect your career goals." Sometimes, the prompt will be as short as, "Tell us something interesting about yourself."
• If there is no prompt, but you still feel the need to introduce your application with a letter, it's usually best to keep it as short as possible. Explain what you're applying for, why you're applying, and thank the contact for their consideration. That's it.

• Often, college prompts will ask you to describe a time you struggled, or a time you overcame some obstacle. Write about something unique, a time that you actually failed and dealt with the consequences.
• The board will get thousands–literally, thousands–of letters about someone's first mission trip, and letters about the time someone's sports team was beaten, then overcame the odds, and won again. Avoid these topics.

• Be specific. If you're writing to a college board, don't say, "I want to go to this college because I need a degree." That's obvious. What do you want to do with it? Why? If you're applying to a business, don't say, "I just need a job." That's obvious. Why this specific job?

• If you're applying to schools, what do you like about the school? What faculty are you interested in? Why this school, instead of another?

Formatting Application Letters

• If you don't get a word-count guideline, just focus on making one or two good points about yourself, and keeping it at that. No need to drone on four several pages.

• Instead of a salutation, write, "Letter of Application" at the top left corner of the page, or put it in the header on the left side at the top.
• If you do have a contact, address it to them, making sure the name is spelled correctly. Then space down and start the body of the letter. [10] X Research source

• Sometimes, it's appropriate to type your name, then print out the letter and sign it in pen. That can be a nice touch.

• Mailing address
• Telephone and/or fax number

Expert Q&A

• Remember to be formal at all times. Do not use abbreviations anywhere. Thanks Helpful 0 Not Helpful 1

You Might Also Like

• ↑ https://www.indeed.com/career-advice/finding-a-job/how-to-write-an-application-letter
• ↑ https://owl.purdue.edu/owl/subject_specific_writing/professional_technical_writing/tone_in_business_writing.html
• ↑ Alexander Ruiz, M.Ed.. Educational Consultant. Expert Interview. 18 June 2020.
• ↑ https://advice.writing.utoronto.ca/types-of-writing/admission-letters/
• ↑ https://wts.indiana.edu/writing-guides/personal-statements-and-application-letters.html
• ↑ https://owl.purdue.edu/owl/job_search_writing/job_search_letters/cover_letters_1_quick_tips/quick_formatting_tips.html
• ↑ https://writing.wisc.edu/handbook/assignments/coverletters/

About This Article

To format an application letter, start by including your name and contact details in the document header. When choosing a greeting, only use one if you know the person's name your writing to. Otherwise, give the document a title, like "Letter of application" at the top of the page. For the body of the letter, aim to write no more than 1 page of single-spaced paragraphs using a standard font. Finally, conclude your letter with a formal greeting like "Sincerely yours." For tips on how to write a job application letter, read on! Did this summary help you? Yes No

• Send fan mail to authors

Did this article help you?

Featured Articles

Trending Articles

Watch Articles

• Terms of Use
• Privacy Policy
• Do Not Sell or Share My Info
• Not Selling Info

wikiHow Tech Help Pro:

Develop the tech skills you need for work and life

• Search Search Please fill out this field.
• Career Planning
• Finding a Job
• Cover Letters

Sample Cover Letter for a Job Application

What Is an Application Letter?

What to include in your application letter, tips for writing a cover letter, cover letter sample and template, email cover letter sample.

• How to Send an Email Application

Frequently Asked Questions (FAQs)

Alex Dos Diaz / The Balance

What's the best way to write a letter to apply for a job? Your letter should detail your specific qualifications for the position and the skills you would bring to the employer. What’s most important is to show the employer that you’re a perfect match for the job.

Your job application letter is an opportunity to highlight your most relevant qualifications and experience. An effective cover letter will enhance your application, showcase your achievements, and increase your chances of landing an interview.

Review what to include in a job application letter, tips for writing that will get your application noticed, and examples of cover letters and email messages to send when applying for a job.

Key Takeaways

• An application letter accompanies a resume and may be uploaded to a job portal, sent via email, or even sent by postal mail, depending on the employer’s requirements.
• Application letters are an ideal way to show your interest in a job and highlight your most relevant skills.
• It’s important to match your letter to the job description and show the employer you have the qualifications they are seeking.

A   letter of application, also known as a  cover letter , is a document sent with your resume to provide additional information about your skills and experience to an employer. Your letter of application is intended to provide detailed information on why you are an ideal candidate for the job.

Your application letter should let the employer know what position you are applying for, what makes you a strong candidate, why they should select you for an interview, and how you will follow up.

Effective application letters explain the reasons for your interest in the specific organization and identify the most relevant skills that qualify you for the job.

Unless an employer specifically requests a job application letter sent by postal mail, most cover letters today are sent by email or attached as a file in an online application tracking system.

As with all cover letters, a job application letter is divided into sections:

• The heading includes your name and contact information.
• A  greeting  addressed to a specific person, if possible.
• The introduction includes why the applicant is writing.
• The body discusses your relevant qualifications and what you have to offer the employer.
• The close thanks the reader and provides contact information and follow-up details.
• Your  signature to end the letter .

Here’s how to ensure your application supports your resume, highlights your most relevant qualifications, and impresses the hiring manager.

Get off to a direct start.  In your first paragraph, explain why you are writing. Mention the job title, company name, and where you found the job listing. While you can also briefly mention why you are a strong candidate, this section should be short and to the point.

Offer something different than what's in your resume. You can make your language a bit more personal than in your resume bullet points, and you can tell a narrative about your work experience and career.

Application letters typically accompany resumes, so your letter should showcase information that your resume doesn't.

Make a good case.  Your first goal with this letter is to progress to the next step: an interview. Your overarching goal, of course, is to get a job offer. Use your application letter to further both causes. Offer details about your experience and background that show why you are a good candidate. How have other jobs prepared you for the position? What would you bring to the role and the company? Use this space to  emphasize your strengths .

Close with all the important details.  Include a thank you at the end of your letter. You can also share your contact information and mention how you will follow up.

This is a sample cover letter.  Download the cover letter template  (compatible with Google Docs and Word Online) or see below for an email sample.

The Balance

John Donaldson 8 Sue Circle Smithtown, CA 08067 909-555-5555 john.donaldson@email.com

September 6, 2023

George Gilhooley LTC Company 87 Delaware Road Hatfield, CA 08065

Dear Mr. Gilhooley,

I am writing to apply for the programmer position advertised in the Times Union. As requested, I enclose my certification, resume, and references.

The role is very appealing to me, and I believe that my strong technical experience and education make me a highly competitive candidate for this position. My key strengths that would support my success in this position include:

• I have successfully designed, developed, and supported live-use applications.
• I strive continually for excellence.
• I provide exceptional contributions to customer service for all customers.

With a BS degree in computer programming, I have a comprehensive understanding of the full lifecycle of software development projects. I also have experience in learning and applying new technologies as appropriate. Please see my resume for additional information on my experience.

I can be reached anytime via email at john.donaldson@email.com or by phone at 909-555-5555.

Thank you for your time and consideration. I look forward to speaking with you about this employment opportunity.

Signature  (only if a hard copy letter)

John Donaldson

The following is a sample email cover letter to send as part of a job application.

Email Application Letter Example

Subject: Colleen Warren - Web Content Manager Position

Dear Hiring Manager,

I'm writing to express my interest in the Web Content Manager position listed on Monster.com. I have experience building large, consumer-focused, health-based content sites. While much of my experience has been in the business world, I understand the social value of this sector, and I am confident that my business experience will be an asset to your organization.

My responsibilities have included the development and management of website editorial voice and style, editorial calendars, and the daily content programming and production for various websites.

I have worked closely with health care professionals and medical editors to provide the best possible information to a consumer audience of patients. I have also helped physicians use their medical content to write user-friendly and easily comprehensible text.

Experience has taught me how to build strong relationships with all departments in an organization. I have the ability to work within a team, as well as cross-team. I can work with web engineers to resolve technical issues and implement technical enhancements. 

I am confident working with development departments to implement design and functional enhancements, monitor site statistics, and conduct search engine optimization.

Thank you for your consideration.

Colleen Warren colleen.warren@email.com 555-123-1234 www.linked.com/colleenwarren

How to Send an Email Application Letter

If sending your cover letter via email, list your name and the job title you are applying for in the  subject line  of the email:

Colleen Warren - Web Content Manager Position

Include your contact information in your email signature but don't list the employer's contact information.

Do you have to write a cover letter when you apply for a job?

Some employers require cover letters. If they do, it will be mentioned in the job posting. Otherwise, it’s optional but it can help your chances of securing an interview. A cover letter gives you a chance to sell yourself to the employer, showcase your qualifications, and explain why you are a perfect candidate for the job.

How can you use a cover letter to show you’re a qualified candidate?

One of the easiest ways to show an employer how you’re qualified for a job is to make a list of the requirements listed in the job posting and match them to your resume . Mention your most relevant qualifications in your cover letter, so the hiring manager can see, at a glance, that you have the credentials they are looking for.

CareerOneStop. " How Do I Write a Cover Letter? "

• Job Acceptance
• Job Acceptance Letter
• Job AD Description
• Job Advertisement
• Job Analysis
• Job Application
• Job Application Cover Letter
• Job Application Letter
• Job Application Letter for Engineer
• Job Application Letter for Executive
• Job Application Letter for Teacher
• Job Application Letters for Assistant
• Job Appointment

FREE Job Application Letter Templates & Examples

Make a Glowing Job Application Letter that Gets You Noticed by the Recruitment Officer by using Template.net’s Free Printable Job Application Letter Templates. Choose from Professionally-written Document Template Samples Online with a Subject, Your Pertinent Information Details, Date, Contact Person, Name, Title, Employer’s Address, Main Content Paragraph, and Salutation that You can Fully Edit, Download of Free, and Print Easily.

Get Access to All  Letters Templates

• Job Appointment Letter
• Job Brochure
• Job Certificate
• Job Checklist
• Job Description
• Job Estimate
• Job Experience
• Job Experience Certificate
• Job Experience Letter
• Job Fair Flyer
• Job Interview
• Job Offer Letter
• Job Opening
• Job Opportunity
• Job Posting
• Job Proposal
• Job Quotation
• Job Recommendation Letter
• Job Recruitment
• Job Reference
• Job Reference Letter
• Job Resignation Letter
• Job Transition Plan
• Job Vacancy
• Coordinator Job Description

Free Job Application Letter Template, Printable, Download or Share via Email

Make sure your job application catches the attention of the hiring manager and increase your chances of getting hired with Template.net’s free printable job application letter templates. Have a short and simple yet strong and professional job application letter, resume, or cover letter examples that you can craft your formal letter to for applying for a small job employment opportunity, teacher, accountant, or any job vacancies for fresh graduates. All template samples contain fillable and editable original content and are downloadable for printing or sharing digitally through email.

Edit Job Application Letter Online for Free and Download

Get different kinds of job application letter examples in short or formal outlines that you can edit online using our document editor tool. Replace or edit the highlighted parts without writing the whole thing to save you time and money. Make your well-written personal letter request for any job opportunity, whether you’re a student, fresher, or professional for an IT engineer, civil engineer, government, bank, or marketing job positions. Download our templates freely in PDF file format.

• Tools Suggest

Application Letter Generator

Unlock your professional potential with our application letter generator, write some details for application letter..

Related Tools

Are you tired of the job application struggle? The Application Letter Generator powered by "Toolbaz" is your passport to effortless success. Just say goodbye to the frustration of writing perfect letters and welcome a world where your applications stand out.

It's time to embark on a journey towards your dream job with confidence. Get ready for a future where your letters open doors to exciting career opportunities and leave a lasting impression.

Application Letter Generator:

This AI Generator is an innovative instrument that transforms how job searchers approach the development of application letters. Individuals may enter their personal information, career ambitions, and applicable certifications into this digital assistant, and it will produce properly formatted application letters.

It is a time-saving approach that simplifies what can sometimes be a difficult procedure and ensures that each application is tailored to the unique job needs. Using this Application Generator allows job seekers to not only speed their application efforts but also improve their chances of leaving a lasting impression on prospective employers, making it a vital resource in the job search process.

Simplify Your Job Application Process:

In the competitive job market, crafting the perfect application letter can be daunting and time-consuming. This Generator offers a solution by streamlining the process. This tool allows you to input your personal details, qualifications, and job preferences, generating personalized, professional application letters.

With tailored letters for each job, your chances of securing interviews increase. This user-friendly platform simplifies the application process, making job hunting less stressful.

Let’s get into it to learn the speciality of this amazing application letter generator!

Key Elements of Application Letter:

A well-crafted application letter typically includes the following key elements:

Contact Information: Include your own contact details (name, address, phone number, email) at the top of the letter.

Date: The date when you are writing the letter.

Recipient's Contact Information: The name, title, and address of the person or organization to whom you are sending the application.

Salutation: A polite and specific greeting, addressing the recipient. If you don't know the name, use a general salutation, such as "Dear Hiring Manager."

Body Paragraphs: This is where you explain your qualifications, skills, and experiences that make you a suitable candidate for the job.

Closing Paragraph: Summarize your interest in the position, express your desire for an interview, and thank the recipient for considering your application.

Closing Salutation: Use a courteous closing like "Sincerely" or "Yours truly," followed by your name.

These seven elements form the core structure of an effective application letter.

Potential Uses of Employing AI Generator:

Exploring the Boundless Potential Uses of Application Letter Generator:

In a rapidly evolving job market, staying ahead requires innovation and efficiency. One groundbreaking tool with immense potential is the AI Letter Generator. Beyond its primary use in simplifying the job application process, this technology boasts several other compelling applications.

Professional Development:

Not just for job seekers, this AI can also benefit professionals seeking career advancement. It aids in crafting compelling cover letters for internal promotions, enhancing prospects within the current workplace.

Career Counseling:

Career advisors and educators can integrate the AI Generator into their services, providing students with a valuable resource to hone their application skills. It simplifies the educational process and prepares students for future success.

HR Optimization:

Human resource departments can streamline their onboarding processes by using AI Generator to create welcoming and customized letters for new hires, thus fostering a positive start to the employment journey.

How To Use This AI Paractically?

Begin by entering your details in the Field Box.

Choose a creativity level from 1 to 10 for optimal results.

Verify your identity by clicking the "recaptcha" button; this step is essential.

Lastly, click "Write" and our intelligent tool will generate outstanding content based on your input.

Key Features To Look Regarding This AI:

When considering an AI focus on customization, template variety, and an intuitive interface for personalized and professional letters. Ensure it offers a built-in spellcheck and grammar tools, simplifying the writing process. Document export options are essential for versatile application submission. Integration with your resume maintains consistency. Real-time editing and cross-device compatibility increase flexibility. Guidance and tips aid those less experienced in application writing. Lastly, prioritize data security and privacy for your personal information.

In summary, select a tool that excels in customization, professionalism, user-friendliness, and features to enhance the quality of your application materials, streamlining your job application process.

Can an Application Letter Generator replace the need for personalized cover letters?

This Generator may simplify the process of drafting cover letters, but it cannot entirely replace the need for personalized ones. Personalized cover letters enable applicants to convey their enthusiasm, cultural alignment, and in-depth knowledge of the company and its goals. They also allow candidates to address specific job requirements and tailor their qualifications accordingly.

While a generated letter can provide a solid foundation, personalized cover letters demonstrate a deeper commitment to the position and can set candidates apart from others. Thus, striking a balance by using both methods is often the most effective approach for job seekers.

Content Writing Tools

Book & story writing.

Poem & Lyrics Writing

Letter writing tools, re-writing tools, more writing tools.

Media Decision US

The 18 Do’s and Don’ts of Cover Letters Every Job Seeker Should Know

Posted: May 8, 2024 | Last updated: May 8, 2024

Most job seekers don’t spend nearly enough time working on their cover letters, assuming that their resume is enough to get them an interview. But when there is competition, a great cover letter can be the difference between getting an interview and getting passed over. 

Your cover letter is your first impression when you’re applying for a new job, and it should be a good one. It’s also an opportunity to show your personality and demonstrate why you’re a perfect fit for the role.

Writing a cover letter can be a daunting task, but you can do a few simple things to make the process easier. Here are some easy do’s and don’ts that can help you write a great cover letter that will impress employers.

Sell Yourself

Like your resume, your cover letter is your chance to brag (professionally) about why they should hire you. Be proud of your skills and accomplishments, and use them to explain why you are the best candidate for the job.

When you sit down to write a cover letter, think about what will grab the hiring manager’s attention and make them want to learn more about you. What can you say about your skills and experience that will set you apart from the other candidates?

If you can, include specific examples of times when you have excelled in a similar role.

Answer the Question: Why Do You Want to Work Here?

You can be more human and personable in your cover letter than in your resume. So be sure to tell the reader why you want the job . This is especially true if you are making a career change or have been out of work for a while. 

Briefly explain your situation so that the hiring manager doesn’t have any questions about why you’re applying. 

For example, you can say something as simple as: “After ten years of working in office administration, I am interested in finding new challenges in the marketing industry.” 

Address How You Meet the Needs of the Organization

There’s a reason most job applications require a resume and a cover letter. A cover letter gives you a chance to communicate with the organization and elaborate on your resume. It’s your opportunity to explain how you meet the organization’s needs and why you should be selected for an interview.  

When writing a cover letter, it’s important to focus on how you can help the company reach its goals. You need to do your research to do this.

Find out the company’s goals and plans for achieving them. Then, craft a cover letter that demonstrates how your skills and experience can help the company succeed. 

You can also use your cover letter to address some of the other job needs that may be difficult to include on your resume. These are things like having a driver’s license and access to a vehicle or details about your availability, such as when you can start.

Personalize Each Letter

Each employer should receive a personalized cover letter, but don’t worry! You can create one or two cover letter templates and tailor them for each job, just like you should do for your resume.

People still expect your cover letter to follow the formal letter format that includes the date, your name and contact information, and the company’s contact information. Be sure to update each cover letter so that it has the correct details and is addressed to the right person. Addressing your cover letter to the wrong person or sending the wrong letter with your resume probably won’t get a second look. 

If you can’t find who to address the letter to, it’s better to use something generic like “hiring manager” or “hiring team” than the wrong name.

Keep it Short

Almost everyone will tell you that your cover letter must be one page. In most cases, this is great advice. Limiting yourself to one page helps you avoid repetition and really focus on what the hiring manager needs to know.

But the truth is, your cover letter should be as long as it needs to be. 

I have been successful in submitting a two-page cover letter in the past. In this case, I was applying for a position that was actually two part-time jobs combined into one full-time job. The two roles were related but required different skills, so there was no way to address them all with a single-page cover letter.

Make sure your cover letter is free of spelling and grammatical errors. Use Grammarly (which is free) to catch spelling errors, grammar mistakes, and other language issues that you may overlook. This attention to detail will show the employer that you are taking the time to make sure that your letter is professional and that you are taking the job seriously. 

Proofreading your own cover letter (and resume) can be difficult because you have likely read it so many times that you no longer see the mistakes. Having someone else take a look at it with fresh eyes can be helpful. In addition, they may be able to offer suggestions for improvements or point out information that is missing.

Get Their Attention Right Away

Almost every cover letter starts in the same boring way: “I am writing to apply for the [position] job at [company].” This does not tell the employer anything about you or why you are qualified for the job. 

Instead, use the first paragraph to grab the employer’s attention and make them want to read more. 

You can do a few things to make your first paragraph truly stand out: 

• Tell them right away why you are qualified for the position. If you have work experience that matches the required qualifications, mention it first. 
• Use strong, active language to engage the employer and show that you are enthusiastic about the position. 
• Talk about your transferable skills, such as those you gained from previous jobs, volunteering, leadership roles, or your side hustle. Use specific examples to demonstrate how you have used these skills in the past and how they will help you succeed in the position you are applying for.

Starting your cover letter with a strong hook will immediately set you apart from other candidates and demonstrate your dedication and enthusiasm for the role.

Use Action Words

Use strong action words on your cover letter, such as: created, managed, oversaw, and implemented. These words will demonstrate your ability to take charge and get things done. Hiring managers are looking for candidates who can take the initiative and get the job done, so make sure to highlight your relevant experience and skills by using descriptive words .

Address Employment Gaps or Potential Concerns

Your cover letter is also an opportunity to explain any gaps in your employment history or to address any concerns that the employer might have about your candidacy. For example, if you took a few years off to raise your children, use your cover letter to explain how this has prepared you to return to the workforce and be an even better employee.

If you are out of work, don’t try to hide it. Employers may eventually discover the truth, so it’s better to be honest with them from the start.

Explain your situation briefly and focus on the positive – what you have been doing to stay busy and how you are excited to put your skills to use in a new role. Honesty is always the best policy, and employers will appreciate your transparency.

Don’t Repeat Your Resume

Now that you know what you should be doing on your cover letter, let’s talk about some of the things you need to avoid. 

Your cover letter is meant to elaborate on your resume, not repeat it. If it doesn’t tell us anything more than your resume already does, why are you even bothering to write one?

Hiring managers don’t want to read the same information twice. They want to see how you can add value to their organization, not just a list of your past accomplishments.

Use your cover letter to talk about your skills and experience in a more natural way. Expand on what you want an employer to know about yourself and your application. 

Don’t Be Negative

If you are applying for a new job, you are either unemployed or underemployed, hate your current job , or are worried that you may be about to lose it. None of these situations are fun to be in, but you can’t let that show in your cover letter. You have to keep it positive!

You want to show the employer that you are excited about the opportunity and are confident in your ability to do the job. 

If you hate your current job, focus on how you are looking for a new challenge and how you believe this job will be a better fit for you. Or, if you are worried you may lose your job, focus on how you are proactive and are already looking for new opportunities. 

Don’t Discuss Why You Need the Job

Everyone knows that you need a job to make money to support yourself and your family. You don’t need to explain this or the details of your specific situation in your cover letter. Mentioning that you are hoping to buy a new house next year doesn’t matter to an employer. 

What does matter to an employer is what you can do for them. They want to know how you will:

• make their company more money
• save them money
• make their company more efficient
• help them to avoid potential problems

In your cover letter, focus on what you can do for the employer, not on what they can do for you. 

Don’t Make Excuses

Making excuses will only draw more attention to your weaknesses or make you sound like a difficult person to work with.

If you don’t meet 100% of the qualifications they are looking for, that’s okay – just don’t point it out! Let them decide if it’s a deal-breaker or if they are willing to train you in that specific area. They might not even notice!

Avoid making excuses for past job experiences or choices that might negatively reflect on you. If you were fired from a job, for example, simply state that the job wasn’t a good fit and move on. Don’t try to justify your actions or make excuses—this will only make you look bad.

Don’t Lie Or Exaggerate

Many people feel the temptation to lie or exaggerate their skills and experience when applying for a new job. Although lying on your application may seem like a harmless way to make yourself look more qualified, it can lead to serious consequences.

When an employer is interested in hiring you, they will conduct a background check and call your references. If you’re caught lying on your job application, you will likely be immediately disqualified. In some cases, you may even be banned from applying to that company in the future.

Lying on your application can also be a form of fraud, which is a crime in many jurisdictions. Depending on the severity of the lie, you could lose your job, be sued, or even be prosecuted for falsifying documents.

Lying or exaggerating about your experience or education can also lead to problems down the road if you are hired for a position based on false information. For example, if you claim you are proficient at using a specific program that you don’t really know much about, you will struggle in your new role. Not being able to do your job will be stressful and raise questions with your employer. Unless you’re a quick learner, you will probably find yourself job searching again within a few months. 

So, the next time you’re tempted to fudge the truth on your application, remember the potential consequences. Be honest on your applications, and you’ll be much better off in the long run.

Don’t Send a Generic Letter

As mentioned, your cover letter should be unique to each employer and job opportunity. Don’t simply copy and paste the same letter for every job application. A few small tweaks are all you need to make your cover letter specific to each job and increase your chances of getting an interview. 

If it’s obvious that you’ve created one cover letter and are using it repeatedly to apply to dozens of jobs, it gives the impression that you don’t really care if you get this job or not – you just want any job. And while that may be true, you don’t want to create any apprehension with an employer. 

Don’t Use Clichés or Slang Terms

Avoid using clichés, slang, and overly casual language when writing a cover letter. Such language can come across as unprofessional and may not convey the message you are trying to get across in the best way possible. 

Clichés include phrases like “I’m a people person” or “I’m a go-getter.” These phrases are overused and do not add anything unique to your letter. 

Using slang can give the impression that you are not taking the process seriously. It can also make it difficult for the reader to understand what you are trying to say. Instead, focus on using clear and concise language, which will get your point across in a way that is both professional and respectful.

While it is important to be friendly and personable in your letter, being too casual can make you seem unprofessional and could hurt your chances of getting the job.

Don’t Include Unnecessary Personal Information

There are a few reasons why you should not include personal information in your cover letter. First, it is not necessary. The employer is only interested in your qualifications and not your personal life.

Second, while it may seem like a good idea to make yourself seem more relatable, including personal information can actually have the opposite effect. It can make you appear unprofessional.

Third, including personal information on your cover letter can be a privacy concern. If an employer knows too much about your personal life, they could potentially use this information against you. For example, if you mention that you have young children, the employer may assume that you will need to take time off for childcare. As a result, you may be passed over in favor of a candidate without the same responsibilities.

Lastly, sharing personal information in your cover letter could also lead to identity theft. If you include your home address or phone number, a savvy thief could use this information to steal your identity. By including personal information in your cover letter, you could be putting yourself at risk.

Overall, you should always err on the side of caution to protect your privacy. Stick to the facts and let your qualifications speak for themselves.

Cover Letters Are Tricky But Beneficial

It can be difficult to strike the right tone in a cover letter. You want to sound enthusiastic and professional without coming across as desperate or pushy. The goal is to show that you’re a good fit for the company, so focus on that. 

If you’re not sure how to get started, plenty of cover letter examples are available online. Just make sure to tailor the letter to the specific company and position you’re applying for, and only include the skills and experience that you actually have.

With these tips, you should have no problem creating a cover letter that will stand out and help you get hired.

Quick Resume Tips

If you want to make a good impression and stand out from the competition, here are 20 resume do’s and don’ts . Following these simple tips, you can be sure that your resume will make a great impression on employers.

Add Your Side Hustle to Your Resume

Job seekers are told they need to stand out if they want to get hired. But how? One of the easiest ways is to include their side hustle on their resumes . Your side hustle is teaching valuable job skills that can make you a stronger candidate. Not mentioning this on your resume or cover letter is a mistake! 

More for You

This is the salary it takes to be considered rich in every state

Which team has the toughest 2024 schedule in the NFL?

Peanuts by Charles Schulz

Donald Trump Lawyers' Mistake-Filled Week

Burger King Is Giving Away Free Whoppers This Weekend

Inside a 16th-Century Normandy Estate That Looks Like Something Out of a Fairy Tale

A psychologist shares the 6 toxic traits of 'highly narcissistic' parents—and how to deal with them

Comedian Andrew Schulz believes Tom Brady's anger toward Jeff Ross for Robert Kraft joke was '100% real'

13 Remote Jobs That Pay Over $200,000 a Year

Donald Trump's Own Words Used Against Him in Court

‘Have we made a terrible mistake?’ My brother put our names on the deed to our 88-year-old mother’s home. What should we do?

Zits by Jeremy Scott and Jim Borgman

I marched at Auschwitz this week. Here's what I say to campus protesters in America.

NBA News: Patrick Beverley gets big suspension

Opinion: The latest political cartoons

13 Trivia Nuggets To Impress And Delight Your Friends

The 16 highest-paying college majors, 5 years after graduation

US Supreme Court rules against Warner Music in copyright damages case

Stock Market Today: Stocks higher on rate cut bets, Dow aims for 40,000

Meet the Voice Behind Randy Travis's New Song

• AI Generator

53 College Application Letter Stock Photos & High-Res Pictures

Browse 53 college application letter photos and images available, or start a new search to explore more photos and images..

Purdue Online Writing Lab Purdue OWL® College of Liberal Arts

Welcome to the Purdue Online Writing Lab

Welcome to the Purdue OWL

This page is brought to you by the OWL at Purdue University. When printing this page, you must include the entire legal notice.

Copyright ©1995-2018 by The Writing Lab & The OWL at Purdue and Purdue University. All rights reserved. This material may not be published, reproduced, broadcast, rewritten, or redistributed without permission. Use of this site constitutes acceptance of our terms and conditions of fair use.

The Online Writing Lab at Purdue University houses writing resources and instructional material, and we provide these as a free service of the Writing Lab at Purdue. Students, members of the community, and users worldwide will find information to assist with many writing projects. Teachers and trainers may use this material for in-class and out-of-class instruction.

The Purdue On-Campus Writing Lab and Purdue Online Writing Lab assist clients in their development as writers—no matter what their skill level—with on-campus consultations, online participation, and community engagement. The Purdue Writing Lab serves the Purdue, West Lafayette, campus and coordinates with local literacy initiatives. The Purdue OWL offers global support through online reference materials and services.

A Message From the Assistant Director of Content Development 

The Purdue OWL® is committed to supporting  students, instructors, and writers by offering a wide range of resources that are developed and revised with them in mind. To do this, the OWL team is always exploring possibilties for a better design, allowing accessibility and user experience to guide our process. As the OWL undergoes some changes, we welcome your feedback and suggestions by email at any time.

Please don't hesitate to contact us via our contact page  if you have any questions or comments.

All the best,

Social Media

Facebook twitter.

A Prototype-Based Neural Network for Image Anomaly Detection and Localization

• Open access
• Published: 08 May 2024
• Volume 56 , article number  169 , ( 2024 )

Cite this article

You have full access to this open access article

• Chao Huang 1 ,
• Zhao Kang 1 &
• Hong Wu 1  

Explore all metrics

Image anomaly detection and localization perform not only image-level anomaly classification but also locate pixel-level anomaly regions. Recently, it has received much research attention due to its wide application in various fields. This paper proposes ProtoAD, a prototype-based neural network for image anomaly detection and localization. First, the patch features of normal images are extracted by a deep network pre-trained on nature images. Then, the prototypes of the normal patch features are learned by non-parametric clustering. Finally, we construct an image anomaly localization network (ProtoAD) by appending the feature extraction network with L 2 feature normalization, a \(1\times 1\) convolutional layer, a channel max-pooling, and a subtraction operation. We use the prototypes as the kernels of the \(1\times 1\) convolutional layer; therefore, our neural network does not need a training phase and can conduct anomaly detection and localization in an end-to-end manner. Extensive experiments on two challenging industrial anomaly detection datasets, MVTec AD and BTAD, demonstrate that ProtoAD achieves competitive performance compared to the state-of-the-art methods with a higher inference speed. The code and pre-trained models are publicly available at https://github.com/98chao/ProtoAD .

Similar content being viewed by others

PaDiM: A Patch Distribution Modeling Framework for Anomaly Detection and Localization

Deep Industrial Image Anomaly Detection: A Survey

Improving the anomaly detection performance of a geometric transform-based convolutional network.

Avoid common mistakes on your manuscript.

1 Introduction

Anomaly detection (AD) [ 1 , 2 ] aims to detect anomalous samples that are deviated from a set of normal samples predefined during training. Traditional image anomaly detection adopts a semantic AD setting [ 3 , 4 , 5 , 6 ], where anomaly samples are from unknown semantic classes different from the one normal samples belong to. Recently, detecting and localizing subtle image anomalies has become an important task in computer vision with various applications, such as anomaly or defect detection in industrial optical inspection [ 7 , 8 ], anomaly detection and localization in video surveillance [ 9 , 10 , 11 ], or anomaly detection in medical images [ 12 , 13 ]. In this setting, anomaly detection determines whether an image contains any anomaly, and anomaly localization, aka anomaly segmentation, localizes the anomalies at the pixel level. This paper focuses on the second setting, especially industrial anomaly detection and localization. Some examples from the MVTec AD dataset [ 8 ] along with predictions by our method are shown in Fig.  1 .

Examples from the MVTec benchmark datasets. From top to bottom: anomaly samples, anomaly mask, and anomaly score maps predicted by our method

In the above applications, anomalous samples are scarce and hard to collect. Therefore, image anomaly detection and localization are often solved with only normal samples. In addition, anomalous regions within images are often subtle (see Fig.  1 ), making image anomaly localization a more challenging task that has not been thoroughly studied compared to image anomaly detection. Recent anomaly localization methods can be roughly categorized into two classes: reconstruction-based methods and OOD-based (out-of-distribution based) methods.

Reconstruction-based methods are mainly based on the assumption that a model trained only on normal images can not reconstruct anomalous images accurately. They reconstruct image as a whole [ 8 , 12 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 ], or reconstruct in the feature space [ 22 , 23 , 24 ]. Then anomaly detection and localization can be performed by measuring the difference between the reconstructed and original ones. This kind of method always needs cumbersome network training.

OOD-based methods evaluate the degree of abnormality for a patch feature by measuring its deviation from a set of normal patch features, which is intrinsically a patch-wise OOD detecting task. Some methods such as PatchSVDD [ 25 ] and CutPaste [ 26 ] learn feature representation by self-supervised learning. On the contrary, some other methods [ 27 , 28 , 29 , 30 ] simply extract features by deep networks pre-trained on natural image datasets such as ImageNet [ 31 ], and achieve promising and even better performances. Since the number of training patches is much larger than that of training images, the inference time and storage increase remarkably. Different strategies have been proposed to tackle this problem. Napoletano et al. [ 27 ] used k-means to learn the dictionary/prototypes for normal patch features, but they evaluated each test patch independently, resulting in high inference time. SPADE [ 28 ] selects k-nearest normal images for patch-wise evaluation based on the global image features, limiting anomaly localization performance. PaDiM [ 29 ] models the normal patches at each position by a multidimensional Gaussian distribution and measures the anomaly by the Mahalanobis distance between a test patch feature and the Gaussian at the same position. However, both SPADE [ 28 ] and PaDiM [ 29 ] are reliant on image alignment. The current state-of-the-art method, PatchCore [ 30 ], uses greedy coreset subsampling to reduce the inference time and storage significantly.

This paper proposes ProtoAD, a prototype-based neural network for image anomaly detection and localization, to improve OOD-based methods’ inference speed. We assume that all normal patch features can be grouped into some prototypes, and abnormal patch features cannot be properly assigned to any of them. Therefore, image anomaly localization can be performed by measuring the deviation of test patch features from the prototypes of normal patch features. First, the patch features of normal images are extracted by a deep network pre-trained on nature images and are L 2-normalized. Then the prototypes of the normalized normal patch features are learned by a non-parametric clustering algorithm. The cosine similarity between two L 2-normalized vectors is equivalent to the dot product between them. Therefore the cosine similarity between a normalized patch feature and a prototype can be implemented by a \(1\times 1\) convolution. Based on this equivalence, we construct an image anomaly localization network (ProtoAD) by appending the feature extraction network with the L 2 feature normalization, a \(1\times 1\) convolutional layer, a channel max-pooling, and a subtraction operation. We use the prototypes as the kernels of the \(1\times 1\) convolutional layer; therefore, our neural network does not need a training phase. Compared with previous OOD-based methods [ 27 , 28 , 29 , 30 ], ProtoAD can perform the anomaly detection and localization in an end-to-end manner, which is more elegant and efficient. Extensive experiments on two challenging industrial anomaly detection datasets, MVTec AD [ 8 ] and BTAD [ 32 ], demonstrate that ProtoAD achieves competitive performance compared to the state-of-the-art methods with a higher inference speed. This advantage of ProtoAD makes it better match the needs of real-world industrial applications.

2 Related Works

2.1 image anomaly localization.

Anomaly detection is an image-level task to determine whether an image contains any anomaly. On the other hand, anomaly localization is more complex to locate anomalies at the pixel level. Here, we only introduce the methods that can be directly applied to image anomaly localization and roughly categorize current methods into two types: reconstruction-based and OOD-based.

Reconstruction-based methods are mainly based on the assumption that a model trained only on normal images can not reconstruct anomalous images accurately, and anomaly detection and localization can be performed by measuring the difference between the reconstructed and original images. Early reconstruction-based methods [ 8 , 12 , 14 , 15 , 17 ] reconstruct image by auto-encoders (AE), variational autoencoders (VAE) or generative adversarial networks (GAN). However, the neural networks have high generalization capacities and can reconstruct anomalies well. Later, different strategies have been proposed to tackle this problem. Different memory-based auto-encoders [ 16 , 18 , 20 ] have been proposed to reconstruct images with features from memory bank to limit the generalization ability. Student-teacher models [ 22 , 23 ] have been used to reconstruct pre-trained deep features. RIAD [ 19 ] randomly removes partial image regions and reconstructs the image by image in-painting. Glance [ 24 ] trains a Global-Net to regress the deep features of cropped patches based on their context. DRAEM [ 21 ] combines a reconstructive sub-network and a discriminative network and trains them in an end-to-end manner on synthetically generated just-out-of-distribution images.

OOD-based methods evaluate the degree of abnormality for a patch feature by measuring its deviation from a set of normal patch features, which is intrinsically a patch-wise OOD detecting task. Some methods such as PatchSVDD [ 25 ] and CutPaste [ 26 ] learn feature representation by self-supervised learning. On the contrary, some other methods [ 27 , 28 , 29 , 30 ] simply extract features by deep networks pre-trained on natural image datasets such as ImageNet [ 31 ], and achieve promising and even better performances. Since the number of training patches is much larger than that of training images, the inference time and storage increase remarkably. Different strategies such as clustering, density estimation, and sampling have been proposed to tackle this problem. Napoletano et al. [ 27 ] learned a dictionary of normal patches from the training set by k-means, and evaluated each patch of a test image by measuring its visual similarity with the k-nearest neighbors in the dictionary. SPADE [ 28 ] compares patch features of a test image with the patch features at the same position of k-nearest normal images selected based on global image features. However, this oversimplified pre-selection strategy will limit the localization performance. PaDiM [ 29 ] models the normal patches at each position by a multidimensional Gaussian distribution and detect anomaly by the Mahalanobis distance between a test patch feature and the Gaussian at the same position. Both SPADE [ 28 ] and PaDiM [ 29 ] are reliant on image alignment. Recently, PatchCore [ 30 ] constructs the memory bank of locally aware patch features by greedy coreset subsampling, and localizes anomaly by measuring the distances of test patch features to their nearest normal patch features in the bank. As a result, PatchCore achieves a new state-of-the-art and significantly reduces the inference time and storage.

Our method is also an OOD-based method with pre-trained deep features but has several differences from the previous works. Our method uses non-parametric clustering instead of k-means in [ 27 ] to learn the prototypes for normal patch features. More importantly, our method can perform anomaly detection and localization by a network in an end-to-end manner, which is more elegant and efficient than the previous methods. Compared to reconstruction-based methods, our network do not need a cumbersome network training phase.

2.2 Clustering Algorithms

Clustering is a type of unsupervised learning task of dividing a set of unlabeled data points into a number of groups such that the data points in the same groups are more similar to each other than they are to the data points in other groups. Clustering provides an abstraction from data points to the clusters, and each cluster can be characterized by a cluster prototype, such as the centroid of a cluster, for further analysis. Clustering algorithms can be roughly divided into four categories: Partition-based cluster, Density-based clustering, Spectral Clustering, and Hierarchical-based clustering.

Partition-based clustering algorithms divide the data into k groups, where k is the predefined number of cluster. The classical algorithms are k–means [ 33 ] and its variations. Although these algorithms are very fast, they need the number of clusters as a parameter and are sensitive to the selection of the initial k centroids.

Density-based clustering defines a cluster as the largest set of densely connected points and can find clusters of arbitrary shapes. DBSCAN [ 34 ] is the most representative algorithm of this class. It has two parameters, radius length \(\epsilon \) and a parameter MinPts . If there are MinPts points in the radius of \(\epsilon \) of a point, it is regarded as a high-density point.

Spectral Clustering [ 35 ] has recently attracted much attention. Most spectral clustering algorithms need to compute the full similarity graph Laplacian matrix and have quadratic complexities, thus severely restricting their application to large data sets.

Hierarchical clustering [ 36 ] is of two types: bottom-up and top-down approaches. In the bottom-up approach (aka agglomerative clustering), each data point starts as a cluster, and the most similar cluster pairs are iteratively merged according to the chosen similarity measure until some stopping criteria are met. In the top-down approach (aka divisive clustering), the clustering begins with a large cluster including all data and recursively breaks down into smaller clusters. Hierarchical clustering produces a clustering tree that provides meaningful ways to interpret data at different levels of granularity. Recently, Sarfraz et al. [ 37 ] proposed FINCH, a high-speed, scalable, and fully parameter-free hierarchical agglomerative clustering algorithm.

An overview of the proposed method. First, the patch features of normal images are extracted by a deep network pre-trained on nature images. Then, the prototypes of the normal patch features are learned by FINCH clustering. For inference, an image anomaly localization network (ProtoAD) is constructed by appending the feature extraction network with the L 2 feature normalization, a \(1\times 1\) convolutional layer, a channel max-pooling (CMP), and a subtraction operation, and anomaly localization is performed in an end-to-end manner

In [ 27 ], k-means is used to learn the prototypes from normal patch features. To avoid choosing the number of clusters ahead, we adopt FINCH to learn the prototypes for normal patch features.

Our method consists of three steps: patch feature extraction, prototype learning, and anomaly detection and localization. An overview of our method is given in Figure. 2 . We describe them sequentially in the following subsection.

3.1 Patch Feature Extraction

Since the features extracted by pre-trained networks have shown their effectiveness for various visual applications including anomaly detection [ 22 , 23 , 27 , 28 , 29 , 30 ], we also adopt deep networks pre-trained on ImageNet dataset [ 31 ] as the feature extractor, and choose the backbone of Wide-ResNet [ 38 ] as the feature extractor following the previous works [ 28 , 29 , 30 ].

ResNet-like deep networks [ 38 , 39 ] include several convolutional stages. The features become more abstract when the stage goes deeper, but their resolution gets lower. Thus, the feature maps from different stages form a feature hierarchy for an input image. Each spatial position of a feature map has a receptive field and corresponds to a patch/region in an input image; therefore, the feature vector at a spatial position of feature maps can be considered as a feature representation for the corresponding image patch. If the feature maps of a stage have a resolution of \(H\times W\) , they contains \(H\times W\) patch features. The deep and abstract features from the ImageNet pre-trained networks are biased towards the ImageNet classification task and are less relevant to the anomaly detection and localization task. Therefore, we adopt the low- and mid-level (stage 1–3) feature representations and combine them as the patch features. Concretely, the feature maps at the higher-level are bilinearly re-scaled to have the same resolution as the lowest level, then the feature maps at different levels are concatenated together for handling multi-scale anomalies. The extracted features are then L 2-normalized where each feature vector is divided by its L 2 norm.

3.2 Prototype Learning

After feature extraction, the prototypes of the L 2-normalized patch features are learned by a clustering algorithm. Then, the prototypes are used in anomaly detection and localization instead of all the normal patch features to reduce the inference time and storage. There are mainly two concerns in choosing a clustering algorithm. First, the number of patch features is much larger than that of training images. For example, each category of MVTec AD dataset has several hundreds of images, while it has several hundreds of thousands of patch features in our implementation. Therefore, the clustering algorithm should be efficient and scalable to large-scale data. Second, most clustering algorithms have some parameters, e.g., the number of clusters or distance thresholds, which can not be well set without a priori knowledge of the data distribution. Thus, these algorithms demand a tedious parameter tuning process to achieve good performance. To meet the requirements of real applications, we adopt FINCH [ 37 ], a high-speed, scalable, and fully parameter-free hierarchical agglomerative clustering algorithm.

The core idea of FINCH is to use the nearest neighbor information of each data point for clustering, which does not need to specify any parameters and has a low computational overhead. Given the integer indices of the first neighbor of each data point, an adjacency matrix is defined according to the following rules:

where \(\kappa _i^1\) symbolizes the first neighbor of data point i . This sparse adjacency matrix specifies a graph where connected data points form clusters. It directly provides clusters without solving a graph segmentation problem. After computing the first partition, FINCH merges the clusters recursively by using cluster means to compute the first neighbor of each cluster until all data points are included in a single cluster or until some stopping criteria is met. In this work, we define the stopping criteria as the number of cluster is less than a threshold and set the threshold to 10,000 to get good results in our experiments. We choose the last partition as the clustering result, and use the mean vectors of clusters as the prototypes of normal patch features.

When the features are L 2-normalized (making the length of a vector to 1), cosine similarity and Euclidean distance between the normalized features are equivalent in the sense of nearest neighbor searching:

where \(L_2()\) is Euclidean distance, \(\textbf{x}_a\) and \(\textbf{x}_b\) are two L 2-normalized feature vectors, and \(\cos \) is cosine similarity. Therefore, we use cosine similarity for clustering and measuring the deviation of test patch features from norm patch features in the next subsection.

3.3 Neural Network for Anomaly Detection and Localization

When a test image passes through the feature extraction network, \(H\times W\) patch features have been extracted. The anomaly score of each patch feature can be computed by measuring its deviation from the prototypes of normal patch features. We compute the anomaly score of a test patch as one minus the cosine similarity between the normalized test patch feature and its nearest prototype. Formally, the anomaly score for the patch at position ( i ,  j ) can be calculated as

where \(\textbf{x}_{ij}\) is the normalized patch feature at position ( i ,  j ), \(\textbf{m}_k\) is the k -th prototype, and \(\cos \) is cosine similarity. In addition, the image-level anomaly score for a test image can be simply computed by maximizing the anomaly scores of all its patch features.

The cosine similarities between a normalized patch feature and a prototype can be computed by a \(1\times 1\) convolution (dot product) between them. Based on this equivalence, we construct a neural network (ProtoAD) for anomaly detection and localization. First, the L 2 feature normalization and a \(1\times 1\) convolutional layer are appended to the feature extraction network, and outputs feature maps of size \(H\times W \times K\) , including the cosine similarities between the \(H\times W\) normalized patch features and all K prototypes. Then, channel max-pooling (CMP) is applied to the feature maps to get the normal score map of \(H\times W\) , including the cosine similarities between the \(H\times W\) normalized patch features and their nearest prototypes. The anomaly score map can be further obtained by computing one minus the normal score map. This process is illustrated by Fig.  3 . Since the spatial resolution of feature maps is lower than that of an input image, we resize the anomaly score map to the resolution of the input image and use a Gaussian filter to smooth it. Finally, anomaly localization can be achieved by thresholding the anomaly score map, and the anomaly score for the test image can be obtained by maximizing the anomaly score map.

We use the prototypes of normal patch features as the kernels of the \(1\times 1\) convolutional layer. Therefore the proposed neural network does not need a training phase. Compared to previous works [ 27 , 28 , 29 , 30 ], our method can perform the anomaly detection and localization in an end-to-end manner, which is more elegant and efficient.

Anomaly detection and localization process of ProtoAD

4 Experiments

4.1 datasets and metrics, 4.1.1 dataset.

MVTec AD dataset [ 8 ] is a real-world industrial defect detection dataset which has become a standard benchmark for evaluating image anomaly detection and localization methods. It has 5354 high-resolution images belonging to 10 objects and 5 texture categories. The images of each category are split into a training and a testing set. Totally, the training set has 3629 normal images, and the test set has 1725 normal and abnormal images of various defects. The ground truth of the test set contains anomaly labels for image-level evaluation and anomaly masks for pixel-level evaluation.

BTAD (BeanTech Anomaly Detection dataset) is a real-world industrial dataset recently released by [ 32 ]. It contains a total of 2830 real-world images of 3 industrial products. The images of each category are split into a defect-free training set and a testing set, supporting evaluation of both anomaly detection and localization.

We follow the split of the two datasets for training and testing.

4.1.2 Evaluation Metrics

AUROC (Area Under the Receiver Operating Characteristic curve) is the most commonly used metric for anomaly detection, which is independent of the threshold. We use image-level AUROC for evaluating the performance of anomaly detection, pixel-level AUROC for anomaly localization. Since the pixel-level AUROC is biased in favor of large anomalies, we also use PRO-score (per-region-overlap) [ 22 ] to evaluate anomaly localization, which weights ground-truth regions of different sizes equally.

4.2 Experimental Setup

We normalize the size of images from all categories of MVTec AD and BTAD dataset to \(256\times 256\) , center crop images to \(224\times 224\) , and do not apply any data augmentation. The backbone of Wide-ResNet50 pre-trained on ImageNet is employed as the feature extractor in our method as in [ 28 , 29 , 30 ]. We define the stopping criteria for FINCH clustering algorithm as the number of clusters is less than 10,000 and choose the last generated partition as the clustering result. For inference, we up-sample the anomaly score map to image size using bilinear interpolation and smooth it with the Gaussian filter with parameter \(\delta =4\) as in [ 29 ]. We implemented our models in Python 3.7 [ 40 ] and PyTorch [ 41 ], and run experiments on NVIDIA GeForce RTX 2080 Ti.

4.3 Results on MVTec AD

4.3.1 comparison with the state-of-the-art.

We compare ProtoAD with the state-of-the-art methods including both the reconstruction and OOD-based methods. The compared reconstruction-based methods include Uninformed students (U-Student) [ 22 ], RIAD [ 19 ], MKD [ 23 ], Glance [ 24 ], DAAD [ 20 ] and DREAM [ 21 ]. And the compared OOD-based methods include SPADE [ 28 ], PatchSVDD (P-SVDD) [ 25 ], CutPaste [ 26 ], PaDiM [ 29 ], and PatchCore (P-Core) [ 30 ]. We directly use their evaluation results if they have been provided.

We report the evaluation results (pixel-level AUROC and PRO-score) for pixel-level anomaly localization on MVTec AD dataset in Tables 1 and 2 respectively. From table 1 , we can see that the OOD-based methods generally achieve better pixel-level AUROC than the reconstruct-based methods. Among the OOD-based methods, the methods using the pre-trained deep features achieve better pixel-level AUROC than the methods based on self-supervised learning. PatchCore achieves the best pixel-level AUROC, PaDiM the second, and the reconstruct-based method DREAM the third. The pixel-level AUROC of our method is very close to those of PaDiM and DREAM. We also notice that our method is more effective on the texture category and achieves the second best AUROC. Table 2 gives the PRO-score results for methods which have used this metric. Among them, Glance achieves the best result, our method is the second best and outperform other OOD-based methods. After all, our method achieves competitive anomaly localization performance to the state-of-the-art methods.

Figure  4 gives qualitative anomaly localization results of our method on MVTec AD dataset. We can see that our method can give accurate pixel-level localization regardless of anomaly region size and type (see supplementary for more qualitative results).

We also report the image-level AUROC results for anomaly detection in Table 3 . PatchCore achieves the best AUROC again, DREAM the second. Our method remains competitive and achieves the third-best AUROC, which is very close to that of DREAM.

4.3.2 Inference Efficiency

Anomaly detection and localization algorithms need high precision and inference speed to match the requirements of real-world applications. Thus, we also report the inference speed of our method and previous OOD-based methods using pre-trained deep features [ 28 , 29 , 30 ] in the table 4 . In the experiments, all the methods adopt Wide-ResNet50 pre-trained on ImageNet as the feature extractor, center-cropped \(224\times 224\) image as input, and run on the same machine with a NVIDIA GeForce RTX 2080 Ti. For PatchCore, we use the implementation provided by the authors, which downsamples the normal patch features via greedy coreset subsampling (PatchCore- \(x\%\) denotes the percentage x of normal patch features are used in inference) and uses faiss [ 42 ] for nearest neighbor retrieval and distance computations. For PaDiM, we make extensive optimization via GPU acceleration. Compared with the previous methods, our model achieves the highest speed, which is 1.2x, 2.7x, and 9.5x faster than PaDiM, PatchCore, and SPADE, respectively. The high inference speed is mainly because our model performs inference in an end-to-end manner, and the main computation added to the feature extraction network is the \(1\times 1\) convolutional layer. Compared to the reconstruct-based methods, our method does not need a cumbersome network training process.

Qualitative anomaly localization results of our method. From top to bottom: abnormal images, ground-truth, and anomaly score maps produced by our method

4.4 Ablation Study

We report ablations studies on the MVTec AD dataset to evaluate the impact of different components of our method on the performance.

4.4.1 Feature Layer Selection

ResNet-like deep networks [ 38 , 39 ] include several convolutional stages. The feature maps from different stages can compose a feature hierarchy for an image. Since the deepest feature maps in the hierarchy are biased towards the ImageNet classification task, we only adopt the features at the low and middle hierarchy levels (stage 1–3) for anomaly detection and localization. Table 5 gives the performance achieved with the features from different levels and their combination. It can be observed that the features from hierarchy level 2 can achieve the best performance among the first three levels, and a combination of the three levels can further improve the performance. Therefore, our method uses the combination of the first three feature levels as the patch feature.

4.4.2 Partition Selection from Clustering Hierarchy

FINCH is a hierarchical agglomerative clustering algorithm. It recursively merges clusters from the bottom up and provides a set of partitions in a hierarchical structure. Each successive partition is a super-set of its preceding partitions, and the number of clusters in it is smaller than those in the preceding partitions. Thus, we need select a partition from the clustering hierarchy as the clustering result.

We report the performance of our method with different partitions, from the second (P2) to the 6-th (P6) partition of FINCH, in Table 6 (see Table 1 in supplementary for more detailed results). We do not include the first partition because it has a huge number of clusters. The results in Table 6 indicate the average performance decreases along with the merging process. This may be because, when the number of clusters gets smaller, clusters are less compact and unsuitable for anomaly detection. On the other hand, if the number of clusters is too large, there are too many prototypes, and the inference time and storage would increase rapidly. We also give the “Best” performance, which FINCH can achieve by selecting the best partition for each category respectively. This best performance is the upper bound that our method can achieve. However, selecting partition based on the average performance (from P2 to P6) or performance for each category (Best) is time-consuming and not suitable for real applications. In our method, we stop FINCH when the number of cluster is less than 10,000 and use the final partition as the clustering result, and give its results in the last line of Table 6 . Our partition selection rule can achieve performance very close to the best one with only a tenth of clusters. Therefore, our method can reach a good trade-off between effectiveness and efficiency.

4.4.3 FINCH vs. K-Means

We compare FINCH clustering algorithm with k-means for the prototype-based anomaly detection. In our method, we choose the partition generated so far by FINCH which having less than 10,000 clusters as the clustering result. For a fair comparison, we set k to 10,000 for k-means. The results in table 7 indicate that the method based on FINCH (the third column) achieves better performance than that based on k-means (the first column). Although it may achieve better performance for k-means by tuning k, it is time-consuming and not feasible for real applications.

4.4.4 Feature Normalization and Cosine Similarity

We also explore the importance of feature normalization for the prototype-based anomaly detection. As shown in Table 7 , k-means with Euclidean distance on the L 2-normalized features (Norm L2) outperforms k-means with Euclidean distance on the original features (L2) in both anomaly detection and anomaly localization and achieves greater improvements in anomaly detection.

When the features are L 2-normalized, cosine similarity and Euclidean distance are equivalent in the sense of nearest neighbor searching. Therefore, we use cosine similarity for clustering and measuring the deviation of test patch features from norm patch features. We further implement cosine similarity with a \(1\times 1\) convolution and append it to the feature extraction network. Therefore inference can be performed in an end-to-end manner.

4.5 Results on BTAD

In Table 8 , we report the results of our method on the BTAD dataset and compare them with those of the SOTA OOD-based method (SPADE, PaDiM, and ProtoAD) and the approaches adopted in [ 32 ]. In [ 32 ], three reconstruction-based methods have been evaluated, auto-encoder (AE) with MSE loss, auto-encoder with MSE and SSIM loss, and Vision-Transformer-based image anomaly detection and localization (VT-ADL). We report the image-level and pixel-level AUROC for each category and their average for all categories. For anomaly detection, ProtoAD achieved the best image-level AUROC. For anomaly localization, ProtoAD achieved the second-best pixel-level AUROC (97.0), very close to the best one (97.4) achieved by PaDiM. These results show our method’s potential to generalize to new anomalous scenarios.

5 Conclusion

We propose ProtoAD, a new OOD-based image anomaly detection and localization method. First, a pre-trained neural network is used to extract features for image patches. Then, a non-parametric clustering algorithm learns the prototypes for normal patch features. Finally, an image anomaly detection and localization network is constructed by appending the feature extraction network with the L 2 feature normalization, a \(1\times 1\) convolutional layer, a channel max-pooling, and a subtraction operation. As a result, ProtoAD does not need a network training process and can conduct anomaly detection and localization in an end-to-end manner. Experimental results on the MVTec AD dataset and the BTAD dataset show that ProtoAD can achieve competitive performance compared to state-of-the-art methods. Furthermore, compared to other OOD-based methods, ProtoAD is more elegant and efficient. And compared to the reconstruct-based methods, ProtoAD does not need a cumbersome network training process. Therefore, it can better meet the requirements of real applications.

Chandola V, Banerjee A, Kumar V (2009) Anomaly detection: a survey. ACM Comput Surv (CSUR) 41(3):1–58

Article   Google Scholar  

Salehi M, Mirzaei H, Hendrycks D, Li Y, Rohban M, Sabokrou M, et al (2022) A unified survey on anomaly, novelty, open-set, and out of-distribution detection: Solutions and future challenges. Trans Mach Learn Res (234)

Ruff L, Vandermeulen R, Goernitz N, Deecke L, Siddiqui SA, Binder A, Müller E, Kloft M (2018) Deep one-class classification. In: International Conference on Machine Learning, pp. 4393–4402

Sabokrou M, Khalooei M, Fathy M, Adeli E (2018) Adversarially learned one-class classifier for novelty detection. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 3379–3388

Golan I, El-Yaniv R (2018) Deep anomaly detection using geometric transformations. Adv Neural Inf Proc Syst 31

Bergman L, Hoshen Y (2020) Classification-based anomaly detection for general data. arXiv:2005.02359

Mei S, Wang Y, Wen G (2018) Automatic fabric defect detection with a multi-scale convolutional denoising autoencoder network model. Sensors 18(4):1064

Bergmann P, Fauser M, Sattlegger D, Steger C (2019) MVTec AD–A comprehensive real-world dataset for unsupervised anomaly detection. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 9592–9600

Sabokrou M, Fayyaz M, Fathy M, Klette R (2017) Deep-cascade: cascading 3D deep neural networks for fast anomaly detection and localization in crowded scenes. IEEE Trans Image Process 26(4):1992–2004

Article   MathSciNet   Google Scholar  

Sabokrou M, Fayyaz M, Fathy M, Moayed Z, Klette R (2018) Deep-anomaly: fully convolutional neural network for fast anomaly detection in crowded scenes. Comput Vis Image Underst 172:88–97

Sabokrou M, Pourreza M, Fayyaz M, Entezari R, Fathy M, Gall J, Adeli E (2018) Avid: adversarial visual irregularity detection. In: Asian Conference on Computer Vision, pp. 488–505

Schlegl T, Seeböck P, Waldstein SM, Schmidt-Erfurth U, Langs G (2017) Unsupervised anomaly detection with generative adversarial networks to guide marker discovery. In: International Conference on Information Processing in Medical Imaging, pp. 146–157

Li Z, Wang C, Han M, Xue Y, Wei W, Li L-J, Fei-Fei L (2018) Thoracic disease identification and localization with limited supervision. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 8290–8299

An J, Cho S (2015) Variational autoencoder based anomaly detection using reconstruction probability. Spec Lect IE 2(1):1–18

Google Scholar  

Bergmann P, Löwe S, Fauser M, Sattlegger D, Steger C (2018) Improving unsupervised defect segmentation by applying structural similarity to autoencoders. arXiv:1807.02011

Gong D, Liu L, Le V, Saha B, Mansour MR, Venkatesh S, Hengel A (2019) Memorizing normality to detect anomaly: memory-augmented deep autoencoder for unsupervised anomaly detection. In: Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 1705–1714

Liu W, Li R, Zheng M, Karanam S, Wu Z, Bhanu B, Radke RJ, Camps O (2020) Towards visually explaining variational autoencoders. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 8642–8651

Park H, Noh J, Ham B (2020) Learning memory-guided normality for anomaly detection. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 14372–14381

Zavrtanik V, Kristan M, Skočaj D (2021) Reconstruction by inpainting for visual anomaly detection. Pattern Recogn 112:107706

Hou J, Zhang Y, Zhong Q, Xie D, Pu S, Zhou H (2021) Divide-and-assemble: learning block-wise memory for unsupervised anomaly detection. In: Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 8791–8800

Zavrtanik V, Kristan M, Skočaj D (2021) Draem-a discriminatively trained reconstruction embedding for surface anomaly detection. In: Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 8330–8339

Bergmann P, Fauser M, Sattlegger D, Steger C (2020) Uninformed students: student-teacher anomaly detection with discriminative latent embeddings. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 4183–4192

Salehi M, Sadjadi N, Baselizadeh S, Rohban MH, Rabiee HR (2021) Multiresolution knowledge distillation for anomaly detection. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 14902–14912

Wang S, Wu L, Cui L, Shen Y (2021) Glancing at the patch: anomaly localization with global and local feature comparison. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 254–263

Yi J, Yoon S (2021) Patch svdd: Patch-level svdd for anomaly detection and segmentation. In: Computer Vision–ACCV 2020: 15th Asian Conference on Computer Vision, Kyoto, Japan, November 30–December 4, 2020, Revised Selected Papers, Part VI 15, pp. 375–390

Li C-L, Sohn K, Yoon J, Pfister T (2021) Cutpaste: self-supervised learning for anomaly detection and localization. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 9664–9674

Napoletano P, Piccoli F, Schettini R (2018) Anomaly detection in nanofibrous materials by CNN-based self-similarity. Sensors 18(1):209

Cohen N, Hoshen Y (2020) Sub-image anomaly detection with deep pyramid correspondences. arXiv:2005.02357

Defard T, Setkov A, Loesch A, Audigier R (2021) Padim: a patch distribution modeling framework for anomaly detection and localization. In: International Conference on Pattern Recognition, pp. 475–489

Roth K, Pemula L, Zepeda J, Schölkopf B, Brox T, Gehler P (2022) Towards total recall in industrial anomaly detection. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 14318–14328

Deng J, Dong W, Socher R, Li L-J, Li K, Fei-Fei L (2009) Imagenet: a large-scale hierarchical image database. In: 2009 IEEE Conference on Computer Vision and Pattern Recognition, pp. 248–255

Mishra P, Verk R, Fornasier D, Piciarelli C, Foresti GL (2021) Vt-adl: a vision transformer network for image anomaly detection and localization. In: 2021 IEEE 30th International Symposium on Industrial Electronics (ISIE), pp. 01–06

Macqueen J (1965) Some methods for classification and analysis of multivariate observations. In: Proceedings of Berkeley Symposium on Mathematical Statistics & Probability

Ester M, Kriegel H-P, Sander J, Xu X, et al (1996) A density-based algorithm for discovering clusters in large spatial databases with noise. In: Kdd, 96, 226–231

Von Luxburg U (2007) A tutorial on spectral clustering. Stat Comput 17(4):395–416

Karypis G, Han E-H, Kumar V (1999) Chameleon: hierarchical clustering using dynamic modeling. Computer 32(8):68–75

Sarfraz S, Sharma V, Stiefelhagen R (2019) Efficient parameter-free clustering using first neighbor relations. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 8934–8943

Zagoruyko S, Komodakis N (2016) Wide residual networks. In: Proceedings of the British Machine Vision Conference, 2016

He K, Zhang X, Ren S, Sun J (2016) Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 770–778

Van RG, Drake F (2009) Python 3 reference manual. CreateSpace, Scotts Valley, p 1593511

Paszke A, Gross S, Massa F, Lerer A, Bradbury J, Chanan G, Killeen T, Lin Z, Gimelshein N, Antiga L, et al (2019) Pytorch: an imperative style, high-performance deep learning library. Adv Neural Inf Proc Syst 32

Johnson J, Douze M, Jégou H (2019) Billion-scale similarity search with gpus. IEEE Trans Big Data 7(3):535–547

Download references

This research was supported by the National Defense Basic Scientific Research Program of China under Grant JCKY2020903B002.

Author information

Authors and affiliations.

School of Computer Science and Engineering, University of Electronic Science and Technology of China, Chengdu, 611731, Sichuan, China

Chao Huang, Zhao Kang & Hong Wu

You can also search for this author in PubMed   Google Scholar

Contributions

All authors contributed to the study’s conception and design. CH performed material preparation, data collection, and analysis. CH and prof. HW wrote the first draft of the manuscript, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Hong Wu .

Ethics declarations

Conflict of interest.

The authors have no relevant financial or non-financial interests to disclose.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary file 1 (pdf 5217 KB)

Rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Huang, C., Kang, Z. & Wu, H. A Prototype-Based Neural Network for Image Anomaly Detection and Localization. Neural Process Lett 56 , 169 (2024). https://doi.org/10.1007/s11063-024-11466-7

Download citation

Accepted : 25 November 2023

Published : 08 May 2024

DOI : https://doi.org/10.1007/s11063-024-11466-7

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Image anomaly detection
• Image anomaly localization
• Non-parametric clustering
• Prototype-based network
• Find a journal
• Publish with us
• Track your research