Crowd counting is still a very challenging task in crowded scenes. The Convolutional Neural Network (CNN) architectures which estimate the density map directly from the input image put up a good performance. While the existing methods mostly use the multi-scale models to widen their networks, we have proposed a very deep network to address the mask. We use the residual block to avoid that too deep network can not converge. Afterwards, we take extensive experiments in three diversity datasets which demonstrate that our method outperforms other state-of-the-art methods. The excellent performance allows our model to be applied not only in counting crowd accurately but also in estimating pedestrian distribution.
展开▼
