tucan9389

It looks like a conversion issue. You can check the conversion calculation on DrawingJointView.swift and definition of aspectFill option. If you have more question, comment below.

## Models - PoseEstimationForMobile(cpm, hourglass) by (edvardHua)[https://github.com/edvardHua] - PoseNet by Google - Pose Estimation by (Fritz)[https://www.fritz.ai/features/pose-estimation.html] - Body Tracking by ARKit3

@hugoliv I havent considered ARKit before. So if you test on ARKit and share it, I'll appreciate it.

Did you try to remove `~/Library/Developer/Xcode/DerivedData` folder and rebuild?

@rohitshirur007 https://github.com/tucan9389/PoseEstimation-CoreML/pull/20 Here is the guide on how to implement the matching algorithm between two different pose joints. If you have more questions, please tell me.

Hi @ardeal, here are some links that might be helpful to start: * QAT: https://www.tensorflow.org/model_optimization/guide/quantization/training * Example: https://www.tensorflow.org/model_optimization/guide/quantization/training_example Please take a look :)

@ardeal Thanks for asking :-) Q1. Typically **yes**. But I recommend checking yourself whether existing FakeQuant or not. But as far as I know, if there is relu after BatchNormalization,...

@ardeal I'll close this issue. Please let me know if you have additional question :)

@mhyeonsoo Thanks for your reporting. Could you give complete script, including `CustomLayer` implementation, if available? (Reproducible and runnable Colab link is also good to communicate.)

@dhruven-god Thanks for reporting. Could you provide fully reproducible script? (colab is recommended or complete python script with tf, tfmot version specification is also fine) If you can share your...