Blazeface Model inference output interpretation

[razor123max]

System information (Please provide as much relevant information as possible)

• Have I written custom code (as opposed to using a stock example script provided in Mediapipe): yes python script
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04, Android 11, iOS 14.4): Linux
• MediaPipe version:
• Bazel version:
• Solution (e.g. FaceMesh, Pose, Holistic): Facemesh and facedetect
• Programming Language and version ( e.g. C++, Python, Java): Python

I am trying to use BlazeFace model as standalone model with Triton inference server, How do i interpret the output of the model and how do i decide the detected face region from output? I see that output0 is regressors float32[1,896,16] and output1 is classificators float32[1,896,1]

[kuaashish]

Hi @razor123max, Please follow this closed issue #1010 and #1010(comment) for your query. Thank you!

[razor123max]

Thank you for your response @kuaashish, from the comments of the closed issue #1010 does not talk about processing the regressors output with some anchor processing similar to SSD, if i am not wrong? if I only run the tflite model exported in ONNX format the values i get are not in 0-1 range aswell (I presume missing self sigmoidal). If some one can explain the anchor point processing part that would be awesome. Thanks.

[kuaashish]

@razor123max,

Are you still looking for resolution on issue?

[github-actions[bot]]

This issue has been marked stale because it has no recent activity since 7 days. It will be closed if no further activity occurs. Thank you.

[github-actions[bot]]

This issue was closed due to lack of activity after being marked stale for past 7 days.

[google-ml-butler[bot]]

Are you satisfied with the resolution of your issue? Yes No