Union feature and RoI feature?

KaihuaTang / Scene-Graph-Benchmark.pytorch

A new codebase for popular Scene Graph Generation methods (2020). Visualization & Scene Graph Extraction on custom images/datasets are provided. It's also a PyTorch implementation of paper “Unbiased Scene Graph Generation from Biased Training CVPR 2020”

MIT License

1.03k stars 228 forks source link

Union feature and RoI feature? #94

Open DiegoD94 opened 3 years ago

DiegoD94 commented 3 years ago

❓ Questions and Help

Hi Kaihua, thanks for the great work. I have several questions regarding the feature extracting. I found in code that both union feature and roi feature are extracted from FPN (when I run relationship inference on custom images), but from the paper seems they are extracted from Faster RCNN, do you have any idea about this? Thanks

KaihuaTang commented 3 years ago

I think you are correct. I shouldn’t call my detector as Faster-RCNN. This project is built on top of https://github.com/facebookresearch/maskrcnn-benchmark. And the detector config I used was called e2e_faster_rcnn_X_101_32x8d_FPN_1x.yaml in their project, so I said that it’s Faster R-CNN with ResNeXt-101-FPN backbone in my paper. Sorry about my bad expression in the paper, but I did cite [28] Feature pyramid networks for object detection as part of our implementation: