Custom Vision: Chapter 2 - Data Annotation

Computer Vision Jul 23, 2021

Here we are, with a new Chapter in our Custom Vision series. Hope you have enjoyed our series of articles on Vision, where we introduced WHAT does a system learns from images and HOW it can replicate human actions.

Often in our implementations, we prefer going with Cloud solutions or pre-built state-of-the-art solutions. But not all use cases can be solved using this. There come times, when you really need to get your hands dirty and do all the hard work to build custom solutions.

Sometimes, this can get really overwhelming.
Don't worry, we are here to help you out in such situations.

As we mentioned in our introductory article, to build a Customized or Hand Crafted AI Solution, a very important step is Data Annotation.

Data Annotation

Data annotation is a process where users assign metadata to the data. This is useful in preparing training datasets for all kinds of supervised learning problems such as object detection, entity recognition, image classification, image segmentation, etc. This process helps us in guiding the machine to learn the right attributes from the data.

In our Custom NLP: Chapter 2 article, we had introduced you to an awesome tool that can be used to annotate your data – Label Studio.

We covered the installation process and how to make use of Label Studio for annotating text data. If you missed the previous article, click here to have a look.

Label Studio can help us in various types of annotation tasks. Here is a list of labelling setups that are available in Label Studio.

  1. Computer Vision
  2. Natural Language Processing
  3. Audio and Video Processing
  4. Time Series Analysis
A really comprehensive list, Isn't it?

Today's consideration will be to make use of Label Studio in the Computer Vision use case. So what are the different annotation tasks related to images, with which the tool can help us?

Enough talking, let's get our hands dirty now. We will try to set up a project on Label Studio for the simplest Computer Vision use case – Image Classification.

Upload Data

We are considering an animal dataset for this demo. You can download the dataset directly from Kaggle. The dataset consists of images of Pandas, Dogs, and Cats. In our demo, we will try to set up Label Studio which will then be used to classify the images into their respective classes.

Sample images from the animal dataset

We will have to create a new project in the Label Studio instance. The type of annotation also needs to be selected from the available options. We will select "Image Classification" and select the data that needs to be uploaded from the local system.

Project creation and Data upload process


With the type of annotation selected and the required data uploaded, we can now get started with the actual data annotation task. On the project home screen, click on the "Label" button on top and start assigning the correct class to the image that pops up on the screen.

Annotation process


With complete data labeled, it's time now to export the annotation information in the required format. There are several file formats available in which the annotations can be exported. This exported data file then can be used to feed data into neural networks which are created. The exported file consists of all the information that is saved as part of the labeling process.


We have seen in this article, what is the use of Image Annotations and how can we annotate an image using Label Studio. There are multiple other tools available in the market which makes the tedious task of data labeling, simpler for you.

Feel free to explore other features available in Label Studio. Also, don't limit yourself to one tool, there are several products available on the market. Explore and find the one which suits your use case.

With the Data Annotation step covered, we can move on to the next step, where we train a simple Neural Network to classify images into different classes. STAY TUNED to get access to awesome content on Computer Vision topics.


Arpit Jain

Machine Learning Engineer

Great! You've successfully subscribed.
Great! Next, complete checkout for full access.
Welcome back! You've successfully signed in.
Success! Your account is fully activated, you now have access to all content.