Did you try a recreate any deep learning experiment from any deep learning paper? It’s a definition of a DEPENDENCY HELL, everything is always broken or falling apart.
Every time you try to install opencv you are not sure if should use channel conda-forge or anaconda, I thought that the package manager
Take a pose estimation for example. 26k stars repo that based its implementation on caffe https://github.com/CMU-Perceptual-Computing-Lab/openpose
If you go to the website https://caffe.berkeleyvision.org/ right now link to the demo is broken.
this is just a rage post…