Plant Disease Detection using Faster R-CNN

Faster R-CNN was first introduced in 2015 and is also a part of the R-CNN family. Compared to its predecessor, Faster R-CNN proposes a novel Region Proposals Network (RPN) and provides better performance and computational efficiency. The whole algorithm can be summarized by merging an RPN (region proposal algorithm) and Fast R-CNN (detection network) into a single network by sharing their convolution features. The paper states that by sharing convolutions at test-times, the cost of computing proposals is as small as 10ms per image. This post gives a brief introduction to the Faster R-CNN algorithm and implementing it to detect leaf diseases.