Create API in order to process video flow on remote GPU server

[TanguyJeanneau]

• Create a local toy API and access it through bash (curl)
• Create a local API that loads and query a model
• Create an endpoint that takes an image path, process the image, then save it (and curl the endpoint)
• Create a .py file that access the endpoint
• Now give the endpoint an image instead of an image path
• Now loop on webcam images
• finally implement the API on a remote GPU server and query it locally
• Fix compression method to avoid the bugged 30s of file decoding on server
• Try to compress / uncompress images before sending and see if speed augment. See https://gist.github.com/andres-fr/f9c0d5993d7e7b36e838744291c26dde