Deep Learning for Computer Vision

Faster R-CNN

Faster R-CNN is proposed by Ren et al. (https://arxiv.org/pdf/1506.01497.pdf). The difference between Faster R-CNN and the Fast R-CNN method is that the Faster R-CNN uses CNN features of architecture such as VGG and Inception for proposals instead of selective search. The CNN features are further passed through the region proposal network. A sliding window is passed through features with potential bounding boxes and scores as the output, as well as a few aspect ratios that are intuitive, the model outputs bounding box and score:

Reproduced with permission from Ren et al.

Faster R-CNN is faster than Fast R-CNN as it saves computation ...