PackagesNotFoundError

[DeXtmL]

trafficstars

Clean Conda installation and Win64, I get this:

PackagesNotFoundError: The following packages are not available from current channels:

• redner-gpu==0.4.25=pypi_0
• tifffile==2020.9.3=pypi_0
• networkx==2.5.1=pypi_0
• pillow==8.4.0=pypi_0
• decorator==4.4.2=pypi_0
• pytorch==1.3.1=py3.6_cuda101_cudnn7_0
• imageio==2.15.0=pypi_0
• scikit-image==0.17.2=pypi_0
• scipy==1.5.4=pypi_0
• face_alignment==1.2.0=py_1
• pyparsing==3.0.7=pypi_0
• torchvision==0.4.2=py36_cu101
• mediapipe==0.8.3

Current channels:

• https://conda.anaconda.org/conda-forge/win-64
• https://conda.anaconda.org/conda-forge/noarch
• https://repo.anaconda.com/pkgs/main/win-64
• https://repo.anaconda.com/pkgs/main/noarch
• https://repo.anaconda.com/pkgs/r/win-64
• https://repo.anaconda.com/pkgs/r/noarch
• https://repo.anaconda.com/pkgs/msys2/win-64
• https://repo.anaconda.com/pkgs/msys2/noarch

[lex3001]

did you ever get this resolved? torchvision==0.4.2 in particular is giving me this error; if i just install it without specifying a version then pytorch will fail

[bboylalu]

I got it running with the commands in the INSTALL file, not the environment.yml, but these were the issues:

• torchvision==0.4.2 is not in the pytorch channel anymore [https://anaconda.org/pytorch/pytorch/files?page=13] (I suppose it was removed). Closest one is: conda install pytorch==1.5.1 torchvision==0.6.1 cudatoolkit=10.1 -c pytorch
• I couldn't install mediapipe because it couldn't build opencv-python, as it tried to install the latest version which dropped support for Python 3.6.* (the one INSTALL mentions). Maybe you can update to Python 3.7, but what I did was: pip install opencv-python==4.6.0.66 which is the last version to support Python 3.6.* and afterwards I installed mediapipe with pip install mediapipe. If you're going for FAN I think this isn't required at all, but maybe it helps someone. Maybe they are not the most elegant solutions, but I've got my *.OBJ file.

[GuyPaddock]

Related to this issue, I was getting this error when trying to install packages:

ERROR: Could not find a version that satisfies the requirement mediapipe (from versions: none)
ERROR: No matching distribution found for mediapipe

It looks like mediapipe no longer publishes a Python 3.6 package, and NextFace uses Python 3.6. The lowest version currently available is for Python 3.8. I changed the INSTALL commands to the following:

conda create --name faceNext python=3.8
conda activate faceNext
conda install pytorch==1.5.1 torchvision==0.6.1 cudatoolkit=10.1 -c pytorch
pip install redner-gpu
conda install -c conda-forge opencv
conda install -c 1adrianb face_alignment=1.2.0
conda install -c anaconda h5py
pip install mediapipe

That gets me past install, but it looks like the way numpy handles delimiters has changed, per https://github.com/numpy/numpy/issues/22311 because I'm getting this error:

Traceback (most recent call last):
  File "optimizer.py", line 489, in <module>
    optimizer = Optimizer(outputDir, config)
  File "optimizer.py", line 21, in __init__
    self.pipeline = Pipeline(self.config)
  File "/home/guyep/playground/NextFace/pipeline.py", line 26, in __init__
    self.morphableModel = MorphableModel(path = config.path,
  File "/home/guyep/playground/NextFace/morphablemodel.py", line 150, in __init__
    self.landmarksAssociation = torch.tensor(np.loadtxt(pathLandmarks, delimiter='\t\t')[:, 1].astype(np.int64)).to(device)
  File "/home/guyep/miniconda3/envs/faceNext/lib/python3.8/site-packages/numpy/lib/npyio.py", line 1356, in loadtxt
    arr = _read(fname, dtype=dtype, comment=comment, delimiter=delimiter,
  File "/home/guyep/miniconda3/envs/faceNext/lib/python3.8/site-packages/numpy/lib/npyio.py", line 999, in _read
    arr = _load_from_filelike(
TypeError: Text reading control character must be a single unicode character or None; but got: '\t\t'