wei-tim / YOWO

You Only Watch Once: A Unified CNN Architecture for Real-Time Spatiotemporal Action Localization
840 stars 161 forks source link

How to get a file similar to ava_categories_ratio.json #78

Open 3sunny opened 3 years ago

3sunny commented 3 years ago

I want to extract 5 classes from AVA dataset for training, but I can't learn information from the ratio file generated according to the number of classes and the total number of classes

lix4 commented 3 years ago

I don't have the exact answer. But I think AVA is multilabel dataset which mean each bounding box might have multiple predicted labels. As a result, we cannot just divide each count by total number.

3sunny commented 3 years ago

Thank you for your reply. Do you know how to write a single category file similar to ava_categories_ratio.json? Is the corresponding ratio set to 1?