Version 2 of the DeepStack Custom Model for object detection API to detect human actions in images and videos. It detects the following actions
- calling
- clapping
- cycling
- dancing
- drinking
- eating
- fighting
- hugging
- kissing
- laughing
- listening-to-music
- running
- sitting
- sleeping
- texting
- using-laptop
Download the model actionnetv2.pt from the Assets section (below) in this release.
This Model is a YOLOv5x DeepStack custom model and that was trained for 150 epochs, generating a best model with the following evaluation result.
mAP@.50: 0.995
mAP@.95: 0.913