Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

RetinaFaceAntiCov TRT Inference

See original GitHub issue

Hello. I’m trying to use this pretrained insightface model with tensorrt. Successfully converted into trt format but having trouble when making predictions. I keep the original retina inference code. Maybe this model has different output shape then it causes an error:

File "/FaceRecognize/src/converters/modules/model_zoo/detectors/retinaface.py", line 322, in postprocess
    landmark_deltas = landmark_deltas.transpose((0, 2, 3, 1)).reshape((-1, 5, landmark_pred_len // 5))
ValueError: cannot reshape array of size 16320 into shape (5,0)

Could you please give me a hand to solve this?

Issue Analytics

State:
Created 3 years ago
Reactions:1
Comments:10 (5 by maintainers)

Top GitHub Comments

1reaction

SthPhoenixcommented, Nov 5, 2020

Hi! RetinaFaceAntiCov has quiet different postprocessing code from original RetinaFace. Also its model has additional output for mask predictions.

You should first add proper output mapping to src/converters/modules/configs.py. Refer to retinaface_outputs as example. New outputs are labeled like face_rpn_type_prob_reshape_stride.

Than you should add proper code for this output processing as in retinaface_cov.py in official repo to src/converters/modules/model_zoo/detectors/retinaface.py

1reaction

xxxpsyduckcommented, Nov 5, 2020

Update: link to converted model