In DeepStream Python, I am working on a face recognition task. The current processing pipeline is pgie (object detection) → tracking → sgie1 (face detection) → sgie2 (face recognition). I am facing an issue where the model continuously detects and recognizes faces (sgie1 + sgie2). However, I want to reduce the computational load by ensuring that faces of already recognized object IDs are not processed again. How can I achieve this?
Do you mean that once you identify a person, you don’t need to identify him in the rest of the video?
Yes, that’s what I mean. Once I identify a person, I don’t need to re-identify them in the rest of the video if the tracking ID remains.
We do not currently support this feature. But you can implement that in your app by yourself.
- Record all the identified tracker ids
- Remove the bbox already recorded in the src_pad of the tracker plugin
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
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.