The code of 'Calibration for Long-tailed Scene Graph Generation'.
Install commond: torch==2.0.1 torchvision==0.15.2. python setup.py build develop. More details on datasets and installations can be found in Scene-Graph-Benchmark.pytorch.
You can evaluate the trained model with the commond: sh scripts/test.sh. The webpath of the pre-trained model is: 'https://drive.google.com/drive/folders/1i6RpbEZOsI1n4KPsj8BAXldHlMzw0MY0?usp=drive_link'
The code is built on Scene-Graph-Benchmark.pytorch, RUNet and PeNet Thanks for their excellent codes.