This repository provides a PyTorch re-implementation of the Cascaded Pyramid Network (CPN), a deep learning architecture that won the MSCOCO keypoints 2017 challenge. The model detects human body keypoints (such as joints) in images to enable multi-person pose estimation. It uses ResNet backbones (ResNet-50/101) with pyramid feature fusion and reports benchmark AP scores on the COCO minival dataset. Training code and evaluation scripts are provided, requiring MSCOCO 2017 dataset annotations.