search

zhihou7 / HOI-CL

Series of work (ECCV2020, CVPR2021, CVPR2021, ECCV2022) about Compositional Learning for Human-Object Interaction Exploration
https://sites.google.com/view/hoi-cl
MIT License
76 stars 11 forks source link

Compositional Splits #21

Closed ASMIftekhar closed 2 years ago

ASMIftekhar commented 2 years ago

Hello, I hope you are doing well. Where can I find the compositional splits you used in this series of papers? Basically, I want to know which images of HICO-DET go into unseen and which images go into seen splits in the testing set?

ASMIftekhar commented 2 years ago

Hello, I think I have found the splits. However, I just want to make sure.

  1. You basically split the training data, but didn't change the test set. Use same test set all along. Am I correct?

  2. Also, I am not sure about zero_shot_type == 7 here, I think you didn't report it anywhere in the paper. Correct me if I am wrong.

https://github.com/zhihou7/FCL/blob/12bc2d4bef4fd2d7d295b33d15f023ae266bc519/lib/ult/ult.py#L398

zhihou7 commented 2 years ago

Sorry for getting you late, I did not check the email recently.

  1. Yes, I did not change the test set.
  2. zero_shot_type ==7 is for the setting when I choose both the tail (24 categories) and head classes (96 categories) for unseen classes. This is for Table 3 in FCL. Sorry for the confusing code.