Car Classification: used ResNet-152 trained on Imagenet, using two branches of convolution only layers for Multiple classification. Used Multiple Loss Functions for Multi-Output Classification Car brand [6 classes] and Car type [5 classes]
Histopathologic Cancer Detection: A LeNet based custom architecture with convolution only layers.
Region Proposal Network is a work in progress project. My attempt to implement the region proposal network part of Faster RCNN from scratch using PyTorch with VGG16 as base network.