How to Build an Object Detection Model using IBM Watson NLC service
Build an Object Detection Model for Brain Tumors without Coding
Object Detection
Object detection is the general form of image classification and localization tasks that fall under Computer Vision and Image Processing. The goal of object detection is to detect specific objects within an image with a confidence score for each object. It is used in different fields for different purposes such as searching, counting, detecting cancer, etc.
Let’s build the model
In this post, we will build an object detection model that detects brain tumors without a single line of code.
After creating a project in IBM Watson, click on the project name and you will be redirected to a similar page as bellow
Click on “add to project” to choose the model type: Visual Recognition Model.
You need to add a service if it is your first time to train or test a visual model
Choose one of the plans that fit your needs
I prefer to keep the default settings. However, you can change the region, plan, service name, etc.
After that, choose one of the visual models (In our case, Detect Objects).
Click on “Browse” to upload your brain scans dataset (or any other images dataset) as a zip file. Shortly after, the images will be loaded to the model where you can see image details.
Click on each image to add objects that you want to detect. Simply, click on “Add objects” button. Draw a square around the object and name it on the right side of the screen. Click “Add” and the object will be added to the image. Finally, click “Done” to save the changes.
Other images may have more than one occurrence of the same object or different object. You can them all. Click “Done” to save the changes.
After adding at least a single object, the model will be ready for training. Click on “Train Model” and the model will be ready in no time.
A notification will be sent after the training is done.
Click “here” to navigate to the model page where you can view all its details.
Click “Test” to inference your model by uploading a new image. The model will detect all possible objects with their confidence scores. In the following example, I have selected to display the tumor objects only.
Finally, click on “Implementation” and will be provided with APIs to inference your model remotely.
