cir7

you can copy and paste it into a text editor, then you will get the secret ak as plaintext.

good idea! @Dai-Wenxun let's try to improve the doc following the suggestion

It seems that some frames of your video don't have any detection results. Currently, the demo can't handle an empty detection result, you could try to fix it.

Thanks for your feedback, it seems that a bug in the ut. We have fixed the bug in https://github.com/open-mmlab/mmaction2/pull/2637, you may have a check.

The average clip operation is performed in the [base head](https://github.com/open-mmlab/mmaction2/blob/582b78fd6c3240500d5cacd292339d7d1ddbb056/mmaction/models/heads/base.py#L196) here. The multiple clips inference result depends on the temporal information richness, if multiple clips share similar information, the improvement...

Sorry for the late response. Generally speaking, multi-clip testing can bring improvement, but if the difference in temporal dimension is small, 8 (frames) x 8 (frame interval) has a span...

I guess that multi-clips inference could be helpful for your project, which could avoid the effect when all frames of a single clip are backgrounds.

@Zepharchit Current skeleton based action recognition models assume that there is only one action in a video/frame, same as the action recognition. what your want to implement is more like...

Hi, @soumyadbanik The demo supports specifying the detection model by assigning `--det-config` and `--det-checkpoint`, and please make sure that the arg `det-cat-id` is assigned as the category id for human...

The warning message is as expected and it's not an error message. The cls_head related weights for your custom dataset is different from the original weights for K400. you can...