# FastRCNN Popular architecture for [[Object Detection]]. Papers: He, Zhang, Ren, Sun, Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition, 2014 Girshick, Fast R-CNN, 2015 Predictions from sliding windows on feature maps (cropping within outputs of layers): ![[fastrcnn.jpg]] ## Region-of-Interest (ROI) pooling How to make sure the dimensionalities match? Divide feature map in TxT cells. The cell size will change depending on the size of the candidate location. ![[roipooling.jpg]] --- ## References 1. 1. Lecture 5.5, UvA DL course 2020