Can 3D action recognition work for non human datasets? I want to classify if a car is moving forward or backwards, and at the same time I want to classify if the car is full or empty so i would have these classes: Forward-full, Forward-empty, Backwards-full, Backwars-empty. Could it work?
Yes,3D action recognition can work for non-human datasets.
Thanks @Morganh. One more question, What if the object is performing the same action under different conditions, for example, car moving forward, but what the car can be loaded or empty. Do you think it is possible to have two classes, i.e. forward-loaded and forward-empty, despite performing the same action, will the model be able to classify them as two different classes?
For forward-loaded and forward-empty, it is suggested to use detection network instead of 3D action recognition.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.