CPU Bottleneck caused by python ? 10% cuda usage

Hello,

I'm using face-alignement in [first-order-model](https://github.com/AliaksandrSiarohin/first-order-model) repo, especially the [crop-video.py](https://github.com/AliaksandrSiarohin/first-order-model/blob/master/crop-video.py) script.

While working on a 1080p video with CUDA, i get this performance :

>32it [00:08, **4.43it/s**]

CPU is used a 54%, CUDA @ 10%

When using CPU mode it's 4 times slower (no surprise here, it seems that CUDA mode is enabled correctly)

>48it [00:44, **1.13it/s**]

Is it a bug from the task manager that doesn't report the correct GPU usage ? Or is it related to python's terrible performances ?

The issue mentions crop-video.py script from first-order-model repo using this library. Start by profiling the script to identify CPU bottlenecks, possibly in data loading or preprocessing. Check CUDA utilization with nvidia-smi and PyTorch profiling tools. Look at the face-alignment library's inference code for potential inefficiencies in GPU-CPU data transfers.

Rédigé par le modèle d'indexation à partir du texte de l'issue.