Action recogition when no action happening

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) - GPU
• DeepStream Version - 6.1.1
• TensorRT Version - 8.4.2.4
• NVIDIA GPU Driver Version (valid for GPU only) - 515

I am running action recognition sample app which is using pretrained model of action recognition which has contained 5 classes [“push”, “fall_floor” , “walk”, “run”, “ride_bike”]. But when there is no action happening or person is missing still getting any one class beacuse of probability, How to handle this case.

Which video did you test with?

i tested with this video

any updates on this issue ?

Sorry for late response! If you are talking about the sample of deepstream-3d-action-recognition sample code, you can modify the postprocessing algorithm in pgie_src_pad_buffer_probe() function to skip label when all probability values are lower than some threshold you set.

how can i set threshold when probalities values are dynamic

The algorithm in the sample code is hardmax. You may apply softmax algorithm to the output classes probabilities, and then you will get the softmax, you can apply the fix threshold to the softmax.

if i set threshold then also it will give any of one class but whenver person is missing in video then how to handle that scenario.

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

you can modify the postprocessing algorithm in pgie_src_pad_buffer_probe() function to skip label when all probability values(after softmax) are lower than some threshold you set.

For example, if you set the threshold to 0.35, and the 5 classes confidence values are 0.2, 0.2, 0.2, 0.2, 0.2, so there is no action, you do not attach any label to the obejct meta.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.