Understanding the relationship between the convolutional feature map and the region proposals in Fast R-CNN

1. What is the difference between R-CNN and fast R-CNN?
2. What is the key idea in faster R-CNN used to increase the speed over fast R-CNN?
3. What is the difference between CNN and R-CNN?
4. How does faster R-CNN work?

What is the difference between R-CNN and fast R-CNN?

This is the basic difference between the Fast R-CNN and Faster R-CNN. Faster R-CNN uses a region proposal method to create the sets of regions. Faster R-CNN possesses an extra CNN for gaining the regional proposal, which we call the regional proposal network.

What is the key idea in faster R-CNN used to increase the speed over fast R-CNN?

Consider using a simple CNN BBox regressor in place of Selective Search to get the approximate region proposals of the image which could further be fed to the underlying Fast R-CNN architecture. This is the core idea behind Faster R-CNN.

What is the difference between CNN and R-CNN?

So, to answer your question: A R-CNN is simply an extension of a CNN with a focus on object detection, while "normal" CNNs are usually used for image classification.

How does faster R-CNN work?

Faster R-CNN is a single-stage model that is trained end-to-end. It uses a novel region proposal network (RPN) for generating region proposals, which save time compared to traditional algorithms like Selective Search. It uses the ROI Pooling layer to extract a fixed-length feature vector from each region proposal.