This repository provides a complete implementation of the Single Shot MultiBox Detector (SSD) algorithm for object detection using PyTorch. It features an object-oriented architecture separating base networks, detection heads, loss functions, and post-processing modules. The project supports multi-GPU training on a single machine and includes methods for detecting objects in both images and videos. All code is documented with Chinese comments for easier understanding.