Implementation of CCNet: Criss-Cross Attention for Semantic Segmentation

Semantic segmentation models, datasets and losses implemented in PyTorch.

PyTorch implementation of DeepLab v2 on COCO-Stuff / PASCAL VOC

I will upload many semantic segmentation models to this repository for you to learn

72.8% MobileNetV2 1.0 model on ImageNet and a spectrum of pre-trained MobileNetV2 models

Classified & clustered every pixel in cityscape dataset image belonging to same object class in modified Google DeepLabV3 architecture. Enhanced boundary detection of the Image Net (ResNet) with atrous conv & Atrous Spatial Pyramid Pooling (ASPP) with variable dilation rates by using end to end Holistic Edge Detection(HED) technique. Added Squeeze & Excitation (SE) block to improve channel attention & also tested softmax in stead of sigmoid for last layer of SE.