Ora-lytics

OCI Data Labeling using Bounding Box

Posted on October 16, 2023 Updated on October 5, 2023

In a previous post, I gave examples of how to label data using OCI Data Labeling. It was a simple approach to data labeling images for input to AI Vision. In that post, we just gave a label for the image to indicate if the image contained a Cat or a Dog. Yes, that’s a very simple approach, and we can build image classification models, and use the resulting model to predict a label for new images. These would be labeled as a Cat or a Dog with a degree of certainty. Although this simple approach can give OK-ish results, we typically want a more detailed model and predictions. For a more detailed approach, we can use Object Detection. For this, we need to prepare our data set in a slightly different way and Yes it does take a bit more time to prepare. Or perhaps it takes a lot more time to prepare the data. But this extra time in preparing the data should (in theory) give us a more accurate model.

This post will focus on creating a new labeled dataset using bounding boxes, and in a later post, we’ll examine the resulting model to see if it gives better or more accurate results.

I’ve mentioned the phrase ‘bounding box’ a few times and this approach does exactly as the phrase indicates. Draw a box around the object and assign a label to it. In our example, we have used a Cats and Dogs dataset. We’ll use that same dataset (50 images of each animal). This approach to labelling the images takes much longer to complete, as we have to draw a box around each animal. But it is worth the effort, as the models can focus on what is inside the box and ignore anything outside the box.

The first task is to define the dataset that will contain our newly labelled data based on the bound box method.

From the OCI menu, go to the Analytics & AI section and select Data Labeling.

Select the Dataset menu items(on the left hand side of the screen, and then click on the Create dataset button.

The Add Dataset screen allows us to enter the details of the dataset we want to use.

Our dataset is based on a dataset in Object storage, and we can define it as the basis of creating a newly labelled dataset. This does not affect the underlying original dataset.

In this case, we want to select Object Detection in the section called Annotation Class.

On the next screen, you can define the Bucket containing the images we want to label.

In our case, we will be using the Cats and Dogs dataset previously loaded into Object Storage.

In the next section, it will tell you how many files are part of the underlying dataset. By default, it will use all of them.

Add the labels you want to use, although you can add more during the labelling process.

Click Next to move to the next screen and then Click Finish to complete this setup.

After a moments, depending on the number of images in the underlying dataset, the version of the dataset for labelling can now be processed.

To stat the labelling process, click on the first on the first image. Using your mouse drag a box over the main item you want to label. In my example, I’m drawing a box around the animals while trying to exclude as much of the surrounding and background parts of the image. After drawing the box, you can then select the label, from the list on the right-hand side of the screen and then click the Save & Next button. Continue doing this until you complete all images. Yes this can take some time, but it should help OCI Vision create a better-informed model for these animals