DeepID-Net: multi-stage and deformable deep convolutional neural networks for object detection

In this paper, we propose multi-stage and deformable deep convolutional neural networks for object detection. This new deep learning object detection diagram has innovations in multiple aspects. In the proposed new deep architecture, a new deformation constrained pooling (defpooling) layer models the deformation of object parts with geometric constraint and penalty. With the proposed multistage training strategy, multiple classifiers are jointly optimized to process samples at different difficulty levels. A ne w pre-training strategy is proposed to learn feature represe ntations more suitable for the object detection task and with good generalization capability. By changing the net structures, training strategies, adding and removing some key components in the detection pipeline, a set of models with large diversity are obtained, which significantly improves the effectiveness of modeling averaging. The proposed approach ranked #2 in ILSVRC 2014. It improves the mean averaged precision obtained by RCNN, which is the stateof-the-art of object detection, from31% to 45%. Detailed component-wise analysis is also provided through extensiv e experimental evaluation.