Can you guess, what is it going to be in this post?Yep, it’s going to be Multi-Class Object Detection as you have seen in the previous posts.
Multi-Class Object Detection
In the previous series, you would have found that our Panda is detected using YoloV3. If you take it as an example, You could identify only Panda and not the other animals.
Well, this Multi-Class detection allows you to train a model and test it with multiple classes under one image, Ex: Panda, Elephant in a picture.
So, this is it – you could see Cat, Dog and bird in a picture.
We will be learning, How to implement Multi-Class Detection in real-time using Detectron 2
Detectron 2 is a next-generation open-source object detection system from Facebook AI Research. With the repository, you can use and train the various state-of-the-art models as the backbone for detection tasks such as bounding-box detection, instance and semantic segmentation, and person key-point identification.
The thing is available as opensource GitHub repo https://github.com/facebookresearch/detectron2
The basic architecture we’re going to use is Faster R-CNN paper is theoretical. If you wish to know more -> Go ahead! However, I’ll be describing it merely.
Base (Faster) R-CNN with Feature Pyramid Network (Base-RCNN-FPN) is a bounding box detector extendable to Mask R-CNN. Faster R-CNN detector with FPN backbone is a multi-scale detector that realizes high accuracy for detecting tiny to large objects, making itself the de-facto standard identifier.
I knew it’s hard to understand.
Let’s get back to our Real-Time Implementation.
Let’s skip the theory and Try to implement Detectron 2 with Faster R-CNN in the next part.