VNext icon indicating copy to clipboard operation
VNext copied to clipboard
verified publisher icon wjf5203

YouTube-VIS and OVIS ablation study splits

Open timmeinhardt opened this issue 3 years ago • 2 comments

In the IDOL paper you run ablation studies on a split of the training set. I also found those in the dataset registration but you did not provide the json files or any script to generate these. Is it possible to share your splits? Thank you very much!

timmeinhardt avatar Aug 17 '22 23:08 timmeinhardt

Thanks for your attention. Here are our splits of YouTube-VIS 2019 and OVIS for oracle experiments in IDOL. https://drive.google.com/drive/folders/1veFOmJzKIr3tgGC8DI-hkBjUg1ivl5tc?usp=sharing

wjf5203 avatar Aug 23 '22 02:08 wjf5203

Thank you for providing those subsets. I hope you can answer another question cause from the code it looks like you used the model trained on pseudo key-reference frame pairs also for OVIS. But in the paper you only mention it for YouTube-VIS 2019.

timmeinhardt avatar Aug 31 '22 15:08 timmeinhardt