This repository provides a complete PyTorch implementation of Faster R-CNN, a two-stage object detector. It supports training on VOC format datasets using ResNet or VGG as backbone networks, with configurable learning rate schedules, optimizers (Adam/SGD), and batch-size-adaptive learning rates. The model achieves around 80.36% mAP@0.5 on VOC07+12 training data using ResNet backbone.