the output of temporal action detection?

[vivva]

I am new to here , I want to know , can the output of the temporal action detection model be the video of the detected action category and boundary ?

[ahkarami]

@dreamerlin , @kennymckormick I think, a demo code for temporal activity detection (for normal & long videos) is required. Best

[kennymckormick]

I think, a demo code for temporal activity detection (for normal & long videos) is required. Best

Thanks for the advice, we will add the item temporal detection demo in the dev plan.

[ZJU-lishuang]

@vivva You can refer this In demo folder ,there already was a SpatioTemporal Action Detection Video Demo

[ahkarami]

@ZJU-lishuang I think Spatio-Temporal action detection differs from just temporal action detection. Can we use the spatio-temporal action detection demo code for the just temporal action detection models (e.g., SSN, BMN) as well?

[ZJU-lishuang]

@ahkarami Where is the the detected action boundary in the temporal action detection?

[ahkarami]

@ZJU-lishuang , Sorry for the late response. Really, I don't understand the meaning of your question precisely. However, I mean that, as you know, the temporal action localization task is different from spatio-temporal action detection task. So, in this view-point their corresponding demo codes would be differ. Currently, in the demo codes, there isn't any specific demo code for temporal action localization task (i.e., demo code related to SSN, BSN, BMN methods). As a result, I advised the team to add a specific demo code for it. Best

[ZJU-lishuang]

@ahkarami I know what you mean. You're right.

[arvindchandel]

@ahkarami Hi, I also want to check Action localization model (SSN, BMN etc.) to generate action proposals. did you run them, if worked then if you can share how to do it?

[ahkarami]

@arvindchandel Hi, Sorry, I've not done that yet.