cv_toy

cv toys / validation on models implementation of different models on mxnet (with python or c++ api).

basic models

generative adversarial networks for hand writing digits generation

basic GAN for hand writing digits from mnist dataset

semantic segmentation

TODO

object detection

Yolo v3

Yolo v3, implemented in MxNet and TensorFlow. TensorFlow version add IoU and GIoU as loss function.

classification

TODO

optical character recognition

TODO

saliency detection

light-weighted saliency detection

Light weighted saliency detection with effort to reduce both modle size and runtime memory, implemented in Pytorch and MxNet.

deepfix

VGG16 backbone and inception module + dilation conv at output layers, implemented in Pytorch

salgan

Generative Adversarial Nets for saliency prediction. Using GANs for saliency prediction is kind of nonesense, but implemented anyway, in Pytorch.