search

yangxuntu / SGAE

220 stars 48 forks source link

more details about image scene graph generation #14

Open chengyj97 opened 4 years ago

chengyj97 commented 4 years ago

Could you please give us more details(codes) about how exactly you generate image scene graphs?

And a lot thanks for your amazing work!

yangxuntu commented 4 years ago

You can just train a simple relation/attribute classifier on VG, which may not decrease the performance much. But you need to extract bottom-up features on VG . The code of extracting features is provided by https://github.com/peteanderson80/bottom-up-attention (caffe version) and https://github.com/zjuchenlong/faster-rcnn.pytorch (pytorch version).

chengyj97 commented 3 years ago

You can just train a simple relation/attribute classifier on VG, which may not decrease the performance much. But you need to extract bottom-up features on VG . The code of extracting features is provided by https://github.com/peteanderson80/bottom-up-attention (caffe version) and https://github.com/zjuchenlong/faster-rcnn.pytorch (pytorch version).

Do you still have the relation/attribute classifier checkpoints? Thanks a lot!

yangxuntu commented 3 years ago

Sorry, I did not have them now. But you can refer to another wonderful scene graph extractor: https://github.com/KaihuaTang/Scene-Graph-Benchmark.pytorch [https://avatars1.githubusercontent.com/u/24402685?s=400&v=4]https://github.com/KaihuaTang/Scene-Graph-Benchmark.pytorch GitHub - 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”https://github.com/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” - KaihuaTang/Scene-Graph-Benchmark.pytorch github.com

发件人: chengyj97 notifications@github.com 发送时间: 2020年12月9日 20:27 收件人: yangxuntu/SGAE SGAE@noreply.github.com 抄送: #YANG XU# S170018@e.ntu.edu.sg; Comment comment@noreply.github.com 主题: Re: [yangxuntu/SGAE] more details about image scene graph generation (#14)

You can just train a simple relation/attribute classifier on VG, which may not decrease the performance much. But you need to extract bottom-up features on VG . The code of extracting features is provided by https://github.com/peteanderson80/bottom-up-attention (caffe version) and https://github.com/zjuchenlong/faster-rcnn.pytorch (pytorch version).

Do you still have the relation/attribute classifier checkpoints? Thanks a lot!

― You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/yangxuntu/SGAE/issues/14#issuecomment-741739339, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AJEJUOSLECJPZWJ4TTSUQGLST5UKVANCNFSM4I637TOA.