The essence of the trick is to refactor each stochastic node into a differentiable function of its parameters and a random variable with fixed distribution.
The result is a continuous-time invertible generative model with unbiased density estimation and one-pass sampling, while allowing unrestricted neural network architectures.
Besides, our experiments converting CIFAR-10 into a point cloud showed that networks built on PointConv can match the performance of convolutional networks in 2D images of a similar structure.
In addition, an ablation study is performed to demonstrate the improvements obtained by different modules in the proposed method.
Estimating crowd count in densely crowded scenes is an extremely challenging task due to non-uniform scale variations.
SOTA for Crowd Counting on UCF CC 50