Comprehensive study of Face R-CNN

• Created the Face R-CNN network from scratch in PyTorch after reviewing, analyzing, and modifying popular object detection models such as Faster R-CNN by introducing a revised loss function and a multi-scale training strategy.
• Examined prominent pre-existing deep learning neural networks such as VGG, LeNet, AlexNet, and FaceNet and learned how they materialized into a new face detection network called Face R-CNN.

Quick Links:

• Report

Results

Results on WIDER FACE dataset. The red boxes or ellipses are the ground-truth annotations. The green boxes are the predicted results. </div>