fatih akyon
fatih akyon
Currently, there is no update. @Rasantis please add a thumbs up emoji in the original post if you would want us to prioritize this feature 🙏🏻
@jersonal, good to hear your interest! We planned to support pretrained video classification models on top of tracking during inference. This wouldn't require high-quality keypoint detections; tracked object boxes would...
@jersonal Hello again; actually, it would be better if you could work on an implementation that utilizes video classification (classifying sequences of frames) instead of classifying a sequence of keypoints....
first step is ready: https://github.com/ultralytics/ultralytics/pull/13935
@IGlace159 you can try it under the examples, itll be added as a new solution in in the following months 👍🏻 https://github.com/ultralytics/ultralytics/blob/main/examples/YOLOv8-Action-Recognition/readme.md
@Muhammad-237 its available as an example, your can try it: https://github.com/ultralytics/ultralytics/tree/main/examples/YOLOv8-Action-Recognition
It works great! However, the auto adjustment doesn't seem to work, and most of the subtitles are out of sync for me.
wanted to say we also need this feature
This feature is already available in the lightning studio (scale to zero). Can you add a simpler version of it to the litserve? cc: @williamFalcon @aniketmaurya
closing this pr in favor of: https://github.com/ultralytics/ultralytics/pull/16441