Yin Guobing

你说的是heat map吧？能力所限，没能复现论文的结果。

This could be helpful: https://github.com/yinguobing/head-pose-estimation

The file `landmark_video.py` is deprecated. If you are targeting inference, try using TensorFlow Serving.

There are two ways to increase the inference speed. 1. using a smaller model. 2. using accelerators like GPU.

iOS automatically handles the model inference process and there is almost nothing we can do on the hardware part. I think using lightweight model like Mobilenet is a better option.

ResNet has a different architecture than Mobilenet, it means you will have to re-train it.

Note that we are not using ResNet in this repository. Theoretically you only need to replace the model in the `model.py` file.

The old project has some issues and that's the why we should update to the latest.

I'm afraid so. The pose is derived from the landmarks. Extreme pose makes it very difficulty to obtain accuracy landmarks.

Hi @soans1994 1. I think you can use branch `features/tfrecord-marks-68` which use the original marks. 2. The only augmentation in this branch is random minor scaling of face areas. However...