KaihuaTang/Scene-Graph-Benchmark.pytorch
A PyTorch codebase for scene graph generation that detects objects and predicts visual relationships between them in images.

This repository provides a PyTorch implementation of scene graph generation methods, including the CVPR 2020 paper on unbiased scene graph generation. It uses Faster R-CNN as a backbone and extends it with scene graph detection capabilities. The toolkit supports object detection, relationship prediction, and graph visualization on both standard benchmarks and custom images. It includes evaluation metrics like recall@K and mean recall@K for assessing scene graph quality.