fix(critical): misc bugfixes to make ROMP usable again

[vivi90]

This merge contains different critical bugfixes to make ROMP usable again.

Some of it are general.
Others are Windows OS related.

Tests

• Windows 10 with Python 3.9.7
• Ubuntu Linux
• Mac OS

Fixed issues

• Error checking compiler version for cl vivi90/ROMP#1
• AttributeError: module 'setuptools._distutils' has no attribute 'version' vivi90/ROMP#2
• ModuleNotFoundError: No module named '...' vivi90/ROMP#3
• FileNotFoundError: [Errno 2] No such file or directory vivi90/ROMP#4
• UserWarning: __floordiv__ is deprecated, and its behavior will change in a future version of pytorch. vivi90/ROMP#5
• UserWarning: Failed to load image Python extension torchvision image vivi90/ROMP#6

Install

choco install visualstudio2019buildtools visualstudio2019-workload-vctools
conda create -n ROMP python=3.9
conda activate ROMP
conda install -n ROMP pytorch=1.10 torchvision=0.11 cudatoolkit=11.3 -c pytorch
pip install -r requirements.txt

Run

python -u -m romp.predict.image --configs_yml=configs/image.yml

Configuration

My configs/image.yml:

ARGS:
 tab: 'process_images'
 GPUS: 0

 backbone: 'hrnet' #'resnet' #
 model_precision: 'fp32' # 'fp16'
 val_batch_size: 4
 nw: 4
 model_path: trained_models/ROMP_HRNet32_V1.pkl #  'trained_models/ROMP_ResNet50_V1.pkl'
 save_visualization_on_img: True
 smpl_mesh_root_align: False
 show_mesh_stand_on_image: False
 soi_camera: 'far' # 'close'
 interactive_vis: False
 renderer: 'pyrender' # 'pyrender' 'pytorch3d' 

 # default: run on demo/images and the results would be saved at demo/images_results
 # for video demo/videos/Messi_1
 inputs: 'E:\Projekte\py\ROMP\demo\images' #'demo/videos/Messi_1'
 collect_subdirs: False # whether to collect images from the sub-folder of the input path.
 output_dir: 'E:\Projekte\py\ROMP\demo\image_results'
 save_mesh: True
 save_centermap: False
 save_dict_results: True
 mesh_cloth: 'ghostwhite'  #'LightCyan'

[Arthur151]

Thanks for your contribution

[Arthur151]

Thanks for your work! I have tested it on Windows. It does fix the bugs of pip whl installation.
I have claimed your contribution at contributor. Is this OK?

[vivi90]

Yes, thank you 🙂