search

luxonis / depthai

DepthAI Python API utilities, examples, and tutorials.
https://docs.luxonis.com
MIT License
946 stars 234 forks source link

Second stage for multiple detection #323

Open InzamamAnwar opened 3 years ago

InzamamAnwar commented 3 years ago

Hi

I would like to know, that, is there any way to apply second stage to all detections in a given frame https://github.com/luxonis/depthai/blob/391bc4c1032ab94b7c31fece6bdf619bbdc3629b/depthai_helpers/mobilenet_ssd_handler.py#L157. For example for emotion recognition or landmark detection, you are only using first face detected to show emotion/landmark. I would like to extract landmarks for all the faces detected

Let me know if you require further details

alex-luxonis commented 3 years ago

On Gen1 we have a limitation in this regard, only first detection was passed to second stage.

Please check this Gen2 demo that was just pushed: https://github.com/luxonis/depthai-experiments/pull/48

VanDavv commented 3 years ago

Using either host side Queue or passthrough output from neural network, you can receive the frame that has been inferenced upon. With https://github.com/luxonis/depthai-experiments/pull/48 I used host-side queue to store the bounding box of a face that I sent to the landmarks neural network, and when the results from this nn arrive, I can consume the bbox to have the full context, with face position, landmarks position on the full frame etc.