Blog

In my previous post What to Desire from a Good Image Annotator?, I wrote about the high-level capabilities of an Image Annotation Tool. In this one, I will go over the requirements for the actual image annotations or as you may also know it, tagging. I will use two images as examples. The first one is a scanned receipt. The receipt example can be used to generalize the broader category of scanned documents, whether financial, legal, or others. The second example is of a cityscape. That one can be used to generalize any other image.

Annotating Store Receipt

Let’s start with the receipt. A receipt is a scanned document that contains financial information. Below is just one way that you may want to annotate a receipt.

Annotated receipt

In this example, I have decided to annotate the receipt using the logical grouping of information printed on it. Each region is a rectangle that contains the part of the image that belongs together. Here is the list of regions and their possible annotations:

  • Region ID: 1
    Annotation: Store Logo
    Description: This can be the store logo or just the name printed on the receipt
  • Region ID: 2
    Annotation: Store Details
    Description: This can include information like address, phone number, store number, etc.
  • Region ID: 3
    Annotation: Receipt Metadata
    Description: This can be the date and time, receipt number as well as another receipt specific metadata
  • Region ID: 4
    Annotation: Cashier Details
    Description: This is information about the cashier
  • Region ID: 5
    Annotation: Items
    Description: Those are the purchased items, the quantities and the individual item price
  • Region ID: 6
    Annotation: Receipt Summary
    Description: This is the summary of the information for the purchase like subtotal amount, tax and the total amount
  • Region ID: 7
    Annotation: Customer Information
    Description: This is information about the customer and any loyalty programs she or he participates to
  • Region ID: 8
    Annotation: Merchant Details
    Description: This is additional information about the merchant
  • Region ID: 9
    Annotation: Transaction Type
    Description: This is information about the transaction
  • Region ID: 10
    Annotation: Transaction Details
    Description: This contains information about the transaction with the payment card processor. It can include transaction ID, the card type and number, timestamp, authorization code, etc.
  • Region ID: 11
    Annotation: Transaction Amounts
    Description: This summarizes the amounts for the transaction with the payment card processor
  • Region ID: 12
    Annotation: Transaction Status
    Description: This is the status of the transaction – i.e., Approved or Declined
  • Region ID: 13
    Annotation: Transaction Info
    Description: Those are technical details about the transaction
  • Region ID: 14
    Annotation: Copy Owner
    Description: This is information about the ownership of the receipt. Usually, this is Merchant or Customer
  • Region ID: 15, 16, and 17
    Annotation: Additional Details
    Description: Those can be various things like return policies, disclaimers, advertisement, surveys, notes, and so on. In this example, we have 15 as Return Policy, 16 as Survey and 17 as Additional Notes

When you think about it, the above areas will be the ones that your eyesd will immediately look to find information. For example, if you want to know what store the receipt was from, you will directly look at the top where the logo should be (Region #1); if you want to know what the total amount is, your eyes will steer towards the receipt summary (Region #6) and so on. Majority of us will follow a similar approach for separating the data because it is something that we do every day in our minds.

Few things to note about the annotations above. First, not every receipt will have all the information from above. Some receipts will have more and some less. Second, annotations evolve. After annotating a certain number of receipts, you start building a pattern and make fewer changes the more you annotate. However, after some time, you may discover that the patterns you developed need to be updated. A straightforward example is a better name for the annotation. If this happens, you need to go back and change the names. Third, there is no standard way to name those annotations. You and I will undoubtedly have different names for the same thing.

Now, let’s write a few requirements from this receipt example.

  1. The first thing we did is to draw the rectangular regions that we want to annotate. And this is our first and simplest requirement.
  2. The second thing we did is to annotate the rectangular region. When we create the annotation, we should be able to add additional information like description of the annotation
  3. The third thing we want is to be able to update annotation information retrospectively.

Those are good as a beginning. But to provide more context and backup our requirements, it will be useful to think about how those annotations will be used, i.e., define our use cases. I kind of hinted to those above.

Use Case #1: Logo Recognition

Let’s say; you are developing classification application that is used to recognize the store the receipt if from. You can easily do this by looking at the store logo only and develop a machine learning algorithm that returns the name of the store by recognizing the logo. For this, the only region you will need is Region 1 with the logo. Thus, you can just cut this region from the receipt and teach your algorithm only on the logo. That way you minimize the noise from the rest of the receipt and your algorithm can have better accuracy.

Use Case #2: Receipt Amount Extraction

If your application needs to extract the summary amounts from the receipt, you can concentrate on Region 6. That region contains all the information you will need. Few things you can do with this region are:

  • Binarize the area
  • Straighten the text
  • OCR the text
  • Analyze the extracted text (not an image related task anymore:))

This use case is applicable for any other are you annotated on the receipt. It doesn’t matter whether you want to obtain the credit card number or the timestamp; the approach will be the same.

Nested Annotations

Now, let’s look at another way to annotate the same receipt.

Annotated Receipt

If your application needs to determine what are your shopping habbits based on geography, you will need to extract detailed information about the store location. Thus, you will want to annotate the receipt as above to know which part is the street address, which is the city, etc. But those regions are all nested in Region 2 from our first annotation pass. It will be useful to have both types of annotations and use them for different use cases.

So, the requirements for the tool will be:

That is also very relevant in the next example, where we have areas with buildings but also want to annotate a single building.

Annotating Cityscapes

Annotating landscapes, cityscapes or other images with real objects is very similar to the receipt annotation. However, real objects rarely have regular shapes in pictures. Here is an example from a picture I took in Tokyo some time ago.

Annotated Cityscape

In this example, I have annotated only a few of the objects: two buildings (1 and 2), a crane (3), soccer field (4) and a tree (5). The requirements for annotating landscapes are not too different from the requirements for annotating documents. There is just one more thing we need to add to the tool to support real object tagging:

There are many use cases that you can develop for real-object recognition, and for that, versatile annotation capabilities will be important in any tool.

Additional Requirements for Annotations

All requirements that I have listed above are specific to the objects or areas in the pictures. However, we need to have the ability to add meta information to the whole picture. Well, you may think we already have a way to do that! We can use the EXIF data. The EXIF data is helpful, and it is automatically populated by the camera or the editing tool. However, it has limited capabilities for free-form meta-information because its fields are standardized.

For example, if you want to capture information who annotated the image last and at what time, you cannot use the EXIF fields for that. You can repurpose some EXIF fields, but you will lose the the original information. What we need is a simple way to create key-value metadata for the image. Of course, having the ability to see the EXIF information would be a helpful feature, although maybe not a high priority one.

With all that, I believe we have enough requirements to start working on tool design. If you are curious to follow the development or participate in it, you can head over to the Image Annotator Github project. The next thing we need to do is to do some design work. That includes UI design, back-end design, and data model.

In my last post, I demonstrated how easy it is to create fake accounts on the major social networks. Now, let’s take a look at what can we do with those fake social network accounts. Also, let’s not forget that my goals here are to penetrate specific individual’s home network (in this case my own family’s home network) and gain access to files and passwords. To do that, I need to find as much information about this individual as possible, so I can discover a weak link to exploit.

1. Search Engines

The first and most obvious way to find anything on the web is, of course, Google. Have you ever Googled yourself? Search for “Toddy Mladenov” on Google yields about 4000 results (at the time of this writing). The first page gives me quite a few useful links:

  • My LinkedIn page
  • Some pages on sys-con.com (must be important if it is so highly ranked by Google). Will check it later
  • Link to my blog
  • Link to my Quora profile
  • Link to a book on Amazon
  • Two different Facebook links (interesting…)
  • Link to a site that can show me my email and phone number (wow! what more do I need 😀)
  • And a link to a site that can show me more about my business profile

The last two links look very intriguing. Whether we want it or not, there are thousands of companies that scrape the Web for information and sell our data for money. Of course, those two links point to such sites. Following those two links I can learn quite a lot about myself:

  • I have (at least) two emails – one with my own domain toddysm.com and one with Gmail. I can get two emails if I register for free on one of the sites or start a free trial on the other (how convenient 😀)
  • I am the CTO co-founder of Agitare Technologies, Inc
  • I get immediate access to my company’s location and phone number (for free and without any registrations)
  • Also, they give me more hints about emails t***@***.com (I can get the full email if I decide to join the free trial)
  • Clicking on the link to the company information additional information about the company – revenue, website as well as names of other employees in the company

OK, enough with those two sites. In just 5 minutes I got enough to start forming a plan. But those links made me curious. What will happen if I search for “Toddy Mladenov’s email address” and “Toddy Mladenov’s phone number” on Google? That doesn’t give me too much but points me to a few sites that do a specialized search like people search and reverse phone lookup. Following some links, I land on truepeoplesearch.com from where I can get my work email address. I also can learn that my real name is Metodi Mladenov. Another site that I learned about recently is pipl.com. A search on it, not surprisingly, returns my home address, my username (not sure for what) as well as the names of relatives and/or friends.

Doing an image search for “Toddy Mladenov” pulls out my picture as well as an old picture of my family.

Let’s do two more simple searches and be done with this. “Metodi Mladenov email address” and “Metodi Mladenov phone number” give my email addresses with Bellevue College and North Seattle College (NSC). I am sure that the NSC email address belongs to the target person because the picture matches. For the Bellevue College one, I can only assume.

Here is useful information I gathered so far with those searches:

  • Legal name
  • 3 email addresses (all work-related, 1 not confirmed)
  • 2 confirmed and 1 potential employer
  • Home address
  • Company name
  • Company address
  • Company phone number
  • Titles (at different employers)
  • Personal web site
  • Social media profiles (Facebook and LinkedIn so far)

Also, I have hints for two personal emails – one on toddysm.com and one on Gmail.

2. Social Media Sites

As you can assume, social media sites are full of information about you. If by now, you didn’t realize it, their business model is to sell that information (it doesn’t matter what Mark Zuckerberg tries to convince you). Now, let’s see what my fake accounts will offer.

Starting with Facebook, I sign in with my fake female account (SMJ) and type the URL to my personal Facebook page that Google returned. What I saw didn’t make me happy at all. Despite my ongoing efforts to lock down my privacy settings on Facebook, it turns out that anyone with a Facebook account can still see quite a lot about me. Mostly it was data from 2016 and before but enough to learn about my family and some friends. Using the fake account, I was able to see pictures of my family and the conversations that I had with friends around the globe. Of course, I was not able to see the list of my friends, but I was still able to click on some names from the conversations and photos and get to their profiles (that is smart, Facebook!). Not only that but Facebook was allowing the fake user to share these posts with other people, i.e., on her timeline (yet another smartie from the Facebook developers).

I was easily able to get to my wife’s profile (same last name, calling me “dear” in posts – easy to figure it out). From the pictures, I was able to see on my and my wife’s profile, I was also able to deduct that we have three daughters and that one of them is on Facebook, although under different last name. Unlike her parents (and like many other millennials), my daughter didn’t protect her list of friends – I was able to see all 43 of them without being her “friend.”

So far Facebook gave me valuable information about my target:

  • Family pictures
  • Wife’s name
  • Teenage daughter’s name
  • A couple of close friends

I also learned that my target and his wife are somehow privacy conscious but thanks to the sloppy privacy practices that Facebook follows, I am still able to gather important intelligence from it.

As we all know, teens are not so much on Facebook but on Instagram (like there is a big difference). Facebook though was nice enough to give me an entry point to continue my research. On to Instagram. My wife’s profile is private. Hence, I was not able to gather any more information about her from Instagram. It was different the case with my daughter – with the fake account I was able to see all her followers and people she follows. Crawling hundreds of followers manually is not something I want to do so, I decided to leave this research for later. I am pretty sure; I will be able to get something more out of her Instagram account. For now, I will just note both Instagram accounts and move on.

Next is LinkedIn. I signed in with the male (RJD) fake account that I created. The first interesting thing that I noticed is that I already had eight people accepting my invite (yes, the one I send by mistake). So, eight people blindly decided to connect with a fake account without a picture, job history details or any other detailed information whatsoever! For almost all of them I was able to see a personal email address, and for some even phone, and eventually address and birth date -pretty useful information, isn’t it? That made me think, how much data can I collect if I build a complete profile.

Getting back on track! Toddy Mladenov’s LinkedIn page gives me the whole work history, associations, and (oops) his birth date. The information confirmed that the emails found previous are most probably valid and can easily be used for a phishing campaign (you see where am I going with this 😀). Now, let’s look what my wife’s page can give me. I can say that I am really proud of her because contact info was completely locked (no email, birthdate or any other sensitive information). Of course, her full resume and work history were freely available. I didn’t expect to see my daughter on LinkedIn, but it turned out she already had a profile there – searching for her name yielded only two results in the US and only one in the area where we live. Doing some crosscheck with the information on Instagram and Facebook made me confident that I found her LinkedIn profile.

3. Government Websites

Government sites expose a lot of information about businesses and their owners. The Secretary of State site allows you to search any company name, and find not only details like legal name, address, and owners but also the owners’ personal addresses, emails and sometimes phone numbers. Department of Revenue web site also gives you the possibility to search for a company by address. Between those two, you can build a pretty good business profile.

Knowing the person’s physical address, the county’s site will give you a lot of details about their property, neighborhood, and neighbors (names are included). Google Maps and Zillow are two more tools that you can use to build a complete physical map of the property – something that you can leverage if you plan to penetrate the WiFi network (I will come back to this in one of my later posts where we will look at how to break into the WiFi network).

Closing Thoughts

I have always assumed that there will be a lot of information about me on the Internet, but until now I didn’t realize the extent to which this information can be used for malicious purposes. The other thing that I realized is that although I tried to follow the guidelines from Facebook and other social media sites to “protect” my information, it is not protected enough, which gives me no confidence that those companies will ever take care of our privacy (doesn’t matter what they say). Like many other people nowadays, I post things on the Internet without realizing the consequences of my actions.

Couple of notes for myself:

  • I need to figure out a way to remove from the web some of the personal information like emails and phone numbers. Not an easy task, especially in the US
  • Facebook, as always fails the privacy check. I have to find a way to hide those older posts from non-friends
  • I also need to educate my family on locking down their social media profiles (well, now that I have that information, I don’t want other people to use it)
  • Hide the birthday from my LinkedIn page

In my next post, I will start preparing a phishing campaign with the goal to discover my home network’s router IP. I will also try to post few suggestions, how you can restrict the access to information from your social media profiles – something that I believe, everyone should do.

Disclaimer: This post describes techniques for online reconnaissance and cyber attacks. The activities described in this post are performed with the consent of the impacted people or entities and are intended to improve the security of those people or entities as well as to bring awareness to the general public of the online threats existing today. Using the above steps without consent from the targeted parties is against the law. The reader of this post is fully responsible for any damage or legal actions against her or him, which may result from actions he or she has taken to perform malicious online activities against other people or entities.

 

 

In my previous post, How Can I Successfully Hack My Home Network? I set the stage for my “Hacking my Home” activities. A possible scenario here is that I am given the task to penetrate a high-profile target’s (i.e., myself 😀) home network and collect as much information to use for malicious purposes. Before I start doing anything though, I need to prepare myself.

Let’s think about what will I need for my activities. Laptop, Internet connection… Well, (almost) everybody has those nowadays. Let’s look at some not so obvious things.

First, I will need some fake online identities that I can use to send emails or connect with people close to my target. For this, of course, I will need to have one or more email addresses. In the past, creating a Gmail or Hotmail address was trivial. Now though, even Yahoo requires you to have a phone number for verification. That makes things a bit more complicated, but not impossible. A quick search on Google or Bing will give you a good list of burner phones that you can buy for as low as $10 at retailers like Fry’s or Wallmart. However, one good option is the Burner app – for just $5/month you can have unlimited calls, texts, and monthly number swaps. There are many uses of it:

  • Use it for general privacy – never give your real phone number on the web or anywhere else (just this is worth the money)
  • Use it for registering new email addresses or social media accounts
  • Make phone calls to avoid Caller ID (we all get those from marketers and scammers)
  • Most importantly, send text messages to avoid Caller ID

If you only need a masked phone, you can also use the Blur app from Abine. There are many other options if you search the web but be careful with the reputation. I decided to try both, Burner and Blur in my next step.

Armed with burner phone numbers, the next step is to create email addresses that I can use for various registrations. Yahoo emails are frowned upon; that is why I decided to go with Gmail. I will need to think of creating a complete online personality that I can use to send emails, engage in conversations, etc. Without knowing what will work, creating a male and female personality will be helpful.

The first step is to choose some trustworthy names. Although fancy, names like North West and Reality Winner 😀 will look more suspicious than traditional English names like John or Mary. Not surprisingly, a simple Google search gives you thousands of pages with names that inspire trustworthiness. So, I choose some of those to name my two personalities. Because I don’t want those to be known, I will use the first initials from now on to differentiate between the male (RJD) and the female (SMJ). Now, let’s go and create some e-mail addresses and social media profiles.

One complication while creating Gmail accounts is the verification process. Google requires your phone number for account creation, and after typing any of the burner phone numbers, I got an error message that the number cannot be used for verification (kind of a bummer, I thought). Interestingly though, after the initial verification (for which I used my real cell phone number), I was able to change the phone number AND verify it using the burner ones.

Next is Facebook. Despite all the efforts Facook is putting to prevent fake accounts, creating a Facebook account doesn’t even require text verification. With just an email, you can easily go and create one. Of course, if you want to change your username, you will be required to verify your account via mobile phone. The burner phones work like a charm with Facebook.

After Facebook, creating an Instagram account is a snap. Picking up a few celebrities to follow is, of course, essential to building up the profile. Be selective when you select people to follow from the suggestion list, you don’t want to look like a bot by following every possible person. Later on, after I gather more intelligence on my targets’ interests, I will tweak the following base to match better their interests, but for now, this will be enough to start building a profile and also getting access to information that is behind the Facebook and Instagram walls.

The Twitter registration process is as simple as Instagram. You don’t need a mobile phone for Twitter – e-mail is enough for the confirmation code. Once again, choose a few select accounts to follow to start building your online profile and reputation.

Creating LinkedIn account is a bit more elaborate. Similar to Gmail, LinkedIn requires real phone number for verification – both burner phones failed here. Of course, LinkedIn doesn’t immediately ask you to add your phone number for notifications. However, their account creation wizard required at least one job title and company. Not unexpected, this stumbled me a bit because I had to think what will be the best title and company to choose from. LinkedIn has one of the cripiest experiences for account creation. The most scary part for me was the ability to pull all my contacts from… well somewhere (I will need to dig into that later on). Nevertheless, my recommendation is to do the registration on a burner laptop that has none of your personal information on it. The next crippy thing was the recommendations – I got some high profile people recommended to me and of course I clicked on a bunch before I realize that it sends invitations to those people. This could have blown my cover because it recommended mostly people working in the company I chose as my employer. Surprisingly to me though just a minute after, I had a request accepted, so I decided that maybe I should not be worried so much. One important thing you need to do on your LinkedIn profile is to change the Profile viewing options in your Privacy settings to Private mode. If you do not do that, people whose profiles you look up will get notified who are you, and you want to avoid that.

Although some people may accuse me in profiling, I have to say that I chose to use the female account for Facebook, Instagram and Twitter and the male for LinkedIn initially. The reason is that I didn’t want to spent time doing all those registrations before I know more about the targets.

Now that I have all those fake accounts created, I can move to the next step: doing some online research about my target. To do that, I don’t need full profiles established but I will try to generate some online activity in the mean time to make those accounts more credible.

In my next post, I will demonstrate how easy it is to gather basic social engineering intelligence using the above accounts as well as free web sites. Stay tuned…

Disclaimer: This post describes techniques for online reconnaissance and cyber attacks. The activities described in this post are performed with the consent of the impacted people or entities and are intended to improve the security of those people or entities as well as to bring awareness to the general public of the online threats existing today. Using the above steps without consent from the targeted parties is against the law. The reader of this post is fully responsible for any damage or legal actions against her or him, which may result from actions he or she has taken to perform malicious online activities against other people or entities.

This morning I was looking at our company’s e-mail gateway and cleaning some of the quarantined messages when I got reminded that while my company’s digital infrastructure may be well protected with firewalls and e-mail gateways, my home network can be wide open and vulnerable to attacks. Like everyone else, I try not to spend too much time configuring my home network and rely on my “ISP to take care of it.” Well, this is a silly approach because the ISPs don’t care about our cybersecurity. It took me hours on the phone, two bricked routers and a house visit (for which I paid of course) to convince mine to replace their outdated router with a simpler gateway device so that I can use my own Eero as the main router and Wi-Fi access point. However, replacing an old router is not something that will solve my cybersecurity issues. Hence, I decided to stop procrastinating and make the first steps to execute on my idea to do some penetration testing on my home network. You will be able to find all my steps and (successful and failing) attempts in the series of Hack My Home post, so let’s get started.

The first thing I need to start with is to decide what my goals are. The best way to do that is to put myself in the hacker’s shoes. If I am a black-hat hacker who wants to attack an Ordinary Joe, what would I like to get from him? Here are a few things that come to mind:

  • Like many of you, I have a file server or NAS device at home, where my family stores a lot of information. Pictures, tax returns, scanned personal documents and what else. Having access to this information may turn beneficial. Hacker’s goal #1: Get access to the file share!
  • Having access to personal information may be useful, but if I am looking for fast money or a way to do bigger damage, harvesting credentials may turn out better. There is a good chance I can find some passwords in a text file on the file share, but because I don’t save mine in plain text, I need to look for other options. Hacker’s goal #2: Steal a password of a family member!

Here is the moment for a disclaimer. Because this is my home, I believe, I have full authority to hack into my devices. If I discover device vulnerability, I will follow the responsible disclosure practice and will need to delay any posts that describe the approach of breaking into the device. Regarding the second goal, stealing a password, I have full (verbal) consent from my family to do that. I also have full access to almost all of their passwords, so I don’t consider this an issue. However, if you are planning to follow my steps, please make sure that you get consent from your family – they may not be so receptive to the idea.

Next, are some assumptions. The biggest one is to assume no knowledge of my home network. Initially, I thought I should start with a diagram of my network, but this will assume I know the details. What I need to do is to get to the details from the outside using public information. If you think about it, the information that hackers can easily (and legally) obtain is the following:

  • Domain name
  • IP address
  • Email address
  • Home address
  • Phone number
  • Social media profiles

This is an excellent set of starting points, isn’t it? Some of those things may be easier obtained than others. Hence, I will need to do some research online to figure out everything I need. I will walk through each step in separate posts. For now, let’s figure out the ways I can digitally break into my home and define some simple next steps.

If I know the IP address of my router, I may be able to attack my home remotely over the Internet. For this though, I need to figure out the IP address of my router. So, one of my next steps would be to figure out an approach to do that.

If I know the location of my home, I may try to attack my home Wi-Fi network and break through it. That will be a little more complicated approach because it will require for me to be close to my home and to use some specialized devices. There may be other wireless devices in my home that I may be able to get to but those will again require some proximity to the home to exploit.

Of course, for my testing purposes, I would like to explore both approaches, but I will need to start with one of them. Because I think the remote exploit higher chance to happen, I would start from there. As a next step, I would need to figure out the entry point for my home from the open Internet, i.e., I need to figure out my router’s IP address.

In my next post, I will walk you through my thought process, the steps and the tools I can use to obtain my home’s IP address.

 

Recently, I started looking for an image annotation tool that we can use to annotate a few thousand images. There are quite a few that you can choose from. Starting with paid ones like Labelbox and Dataturks, to free and open source ones like coco-annotator, imglab and labelimg. The paid ones have also a free tier that is limited but can give you a sense of their capabilities.

Each one of those tools can be used to create an image dataset, but without going into details in each one of them, here are the things we were looking to have to consider an image annotation tool fit for our needs:

  • First, and foremost – usability. I like nice UI, but it shouldn’t take too much time and clicks to get to the core functionality of the tool – i.e., annotating images. Navigating within the annotator, creating projects and adding images should be smooth and easy. Users shouldn’t be tech savvy to set up and start using the annotation tool.
  • Second is security. I wouldn’t put this one so high on the list if I didn’t need to deal with sensitive data. Unfortunately, even some of the commercially available annotation tools neglected basic security features like encryption. Annotating healthcare images like x-rays or financial documents on such platforms will be out of the question. Also, detailed privacy and data handling policies are crucial for handling sensitive images.
  • The third is scalability. The stand-alone annotation tools are limited to the resource of the machine they are running on, but the hosted ones lack detailed information about the limits they impose on the various plans. In addition to the number of projects and users, which are most often quoted in the commercial plans, the number of images per dataset or the total amount of storage available will be good to know.
  • Fourth is versatility. And, with versatility, I don’t mean only the ways one annotates images (polygons vs. rectangles for example) but also export formats like COCO, PASCAL VOC, YOLO, etc.; ability to choose different storage backends like files, AWS S3, Azure Storage or Dropbox and so on. Like everything else in technology, there is no single standard, and a good image annotation tool should satisfy the needs of different audiences.
  • The fifth is collaboration. An image annotator should allow many (if not thousands of) people to collaborate on the project. Datasets like PASCAL VOC or COCO consists of hundreds of thousands of images and millions of annotations, work that is beyond the scale of a single person or small team. Enabling lightweight access to external collaborators is crucial to the success of such a tool.

OK! Let’s be a little bit more concrete about what requirements I have for an image annotation tool.

Image Annotation Tools Usability

Before we go into the usability requirements, let’s look at the typical steps a user must go through to start annotating images:

  1. Select the tool or the service
  2. Install the tool or sign up to the service
  3. Create a project or data set
  4. Chose data storage location (if applicable)
  5. Determine users and user access (if applicable)
  6. Upload image
  7. Annotate image

To select the tool and the service, users need to be clear about what they are getting. Providing a good description of the features the annotation tool offers is crucial for the successful selection. Some developers rely on trials, but I would like to save my time for registering and installation if I know upfront the tool will not work for me.

If the tool is a stand-alone tool, its installation should not be a hassle. A lot of the open source tools rely on the tech savviness of their users, which can be an obstacle. Besides, I am reluctant installing something on my machine if it will turn out not what I need. Trials and easy removal instructions are crucial.

One of the biggest usability issues I saw in the tools are the convoluted flows for creating projects and datasets. Either bugs or unclear flows resulted in a lot of frustration when trying out some of the tools. For me it is simple – it should follow the ages old concept of files and folders.

Being clear where the data is stored and how to access it or get it out of there is important. For some of the tools I have tested, it took a while to understand where the data is; others used some proprietary structures and (surprisingly) no export capabilities.

Adding users and determining access is always a hassle (and not only in image annotation tools). Still, there should be a clear workflow for doing that as well as opening the dataset to the public if needed.

Although I may have an existing dataset, I may want to add new images to it – either one by one or in bulk. This should be one of the most prominent flows in the annotation tool. At any point in the UI I should be able to easily upload a single or multiple files.

For me, the annotation UI should mimic the UI of the traditional image editing software like Adobe Photoshop. You have menus on the top, tools on the left, working are in the middle and properties on the right. Well, it may be boring or not modern but it is familiar and intuitive.

Securing Annotated Images

We deal with scanned financial documents that can contain highly sensitive information like names, addresses, some times credit card details, account numbers or even social security numbers. Some of our customers would like to have a tool that allows them to annotate medical images like x-rays – those images can also contain personal information in their metadata (if, for example DICOM format is used).

Unless the annotation tool is a standalone tool that people can install on their local machine, using secure HTTPS is a no-brainer and the least you can do from security point of view (surprisingly some of the SaaS services lacked even here). However, security goes far beyond that. Things that should be added are:

  • Encrypting the storage where the annotated images are stored. Hosted or self-managed keys should be allowed.
  • Proper authentication mechanisms should be added. Multi-Factor-Authentication should be used for higher security.
  • Good Role Based Access Control (RBAC) should be implemented. For example some people should be able to just view the annotated images, while others to annotate and edit those.
  • Change logs should be kept as part of the application. For example, will be important to know who certain annotation and whether it was correct or not.

Scalability for Image Annotators

A good dataset can contain hundreds of thousands of images – the COCO dataset for 2017 has 118K images in its training dataset. Depending on the quality of the images, the storage needed to store those can vary from 10s of GB to 100s of GB to PB and more. Having an ability to grow the storage is essential to the success of an image annotation tool.

On the users side, a dataset of hundred thousand images may require hundreds of people to annotate. Being able to support large userbase without a huge impact on the cost is also important (hence, the user-based licensing may not be the best option for such a SaaS offering because a single user may annotate only 2-3 images from the whole dataset).

The back-end or APIs handling the annotation of the images should also be able to scale to the number of images and users without problems.

Versatile Export Options for Annotated Images

Rarely the image annotation tool is tightly coupled with the machine learning system that will use the images. Also, the same annotations can be used by various teams using different systems to create the machine learning models. Clear explanation of the format used to store the annotations is a must-have but also the ability to export the annotations in common formats will be essential for the success and usefulness of the tool.

The word “export” hear may mean different things. It doesn’t always need to be download the images and annotations in the desired format but simply saving the annotations in this format.

I would start with defining a versatile format for storing the image annotations and then offer different “export” options, whether for download or just conversion in the original storage.

Collaborating While Annotating Images

Having a single person create an image dataset with hundreds of thousands of images is unrealistic. Such a task requires the collaboration of many people who can be spread around the world. Having the ability to not only give them access to annotate the images but also to comment and give suggestions to already existing annotations is a feature that should be high on the priority list.

Annotations, like software, are not free of bugs. Hence, the image annotation tool should allow for collaboration similar to what modern software development tools enable. This may not be V1 feature but should certainly come soon after.

Now, that I have a good idea what we would like to have from an image annotation tool, it is time to think of how to implement one that incorporates the above mentioned functionality. In the next post, I will look at what we would like to annotate and how to approach the data model for annotations.