So I am aiming to load a BodyPix model offline, however I am having trouble understanding what is needed to do that. I have already successfully used the bodypix model from tensorflow.js to do what I needed online. However, when I looked up what is needed to work offline I came across downloading and locally hosting a .json and .bin files. I understand the reasoning behind them, however aside from links given
json: https://storage.googleapis.com/tfjs-models/savedmodel/bodypix/resnet50/float/model-stride16.json
.bin: https://storage.googleapis.com/tfjs-models/savedmodel/bodypix/resnet50/float/group1-shard1of23.bin
I don't really understand how this works. I understand that I need the weights from the .bin file and the .json should contain information about the model but I want an explanation. I know about the different parameters the model takes in. I did try to search the BodyPix github but I only found typescript files and no .bin files. I also had another problem with webgl not being able to handle images that were processed by bodypix.
I got errors like: Uncaught (in promise) Error: Requested texture size [17344x17344] greater than WebGL maximum on this browser / GPU [16384x16384]
So my questions are:
1. What do I need to run a bodypix model offline? How do I do this?
2. How do people know these links work or where can I find the proper files?
3. How can I avoid the restrictions on image sizes for Webgl?
Side note (I don't plan on using html/js and therefore Webgl to run offline but I want to work one step at a time)
