search

natanielruiz / deep-head-pose

:fire::fire: Deep Learning Head Pose Estimation using PyTorch.
Other
1.57k stars 367 forks source link

I created an executable repo for L2 distance #107

Open MitchellX opened 3 years ago

MitchellX commented 3 years ago

Hi there, I created an executable repo at https://github.com/MitchellX/deep-head-pose. The pre-models are faithful to the author's great work, where there are some bugs and I have fixed them now.

Additionally, I have written some cool scripts in my forked repo. Therefore, you could easily get the [yaw, pitch, roll] results whatever the input you give. For instance, you can give a single image or an image folder! Plus, you can easily use my script to calculate the head pose difference (or so-called L2 distance) between two images or image sets.

Fodark commented 3 years ago

Thanks Mingcan for yours version! I was interested in using your version because it provides an easy way to compute the L2 distance between poses, but unfortunately I am unable to run it on my machine, most probably because I have a 3080 GPU and the torch required both by you and the original author is not supported on this card. Do you have any clue on how to run it on more modern Python+PyTorch version?

MitchellX commented 3 years ago

@natanielruiz Hi NIcola, Please make sure you use the right version python==2.7, torch==0.4.1. Here I recommend virtualenv and conda to create a separate environment.

If all of these well done, but you still can not run the script. Please check out your CUDA version.

Fodark commented 3 years ago

@MitchellX thanks for the reply, indeed it is a CUDA problem, as my 3080 supports only torch 1.7 and later, older version of torch do not work, unfortunately. I'll have to find another way. Thank you nevertheless!