search

cvlab-stonybrook / Scanpath_Prediction

Predicting Goal-directed Human Attention Using Inverse Reinforcement Learning (CVPR2020)
MIT License
97 stars 22 forks source link

Cocosearch18 test set #22

Closed StoyanVenDimitrov closed 2 years ago

StoyanVenDimitrov commented 2 years ago

Hi, I would like to reproduce your results on the Cocosearch18 test set. Do you have the DCB tensors for it, similar to the train and validation datasets, as well as the test set annotations?

ouyangzhibo commented 2 years ago

Please find the all DCB tensors and the bounding box annotations (including the test set) at https://drive.google.com/drive/folders/1C2kK4dGgVWZOgplIjleB5WG18X2yBtRw?usp=sharing.

Note: you can use extract_DCBs_demo.py to extract DCB representation and the bounding box annotation can be extracted using cocoapi.

StoyanVenDimitrov commented 2 years ago

Thank you! I guess the test annotations are here under COCO-Search18 TP Dataset /Testing Set?

ouyangzhibo commented 2 years ago

Thank you! I guess the test annotations are here under COCO-Search18 TP Dataset /Testing Set?

Oh, yes. I thought you mean the bounding box annotations only. I am closing this issue, but feel free to reopen it if there are any further questions.

StoyanVenDimitrov commented 2 years ago

Thank you, I was able to generate test trajectories. But please refer to my comments on #16 . I was not able to reproduce your test results