I've prepared my custom dataset as AVA V2.2 structure. However when I run the train script with ava_v2.2 param an error showed up from ava_helper.py although my dataset structure is AVA v2.2. Is there any document that explain what is the convenient data structure for training as multi-person spatio-temporal action detection ?
Hello,
I've prepared my custom dataset as AVA V2.2 structure. However when I run the train script with ava_v2.2 param an error showed up from
ava_helper.py
although my dataset structure is AVA v2.2. Is there any document that explain what is the convenient data structure for training as multi-person spatio-temporal action detection ?Best Alper