Node-Python ML Image Classification Model Boilerplate

Starter code to use NodeJS with a Python layer for the model. Useful for deploying deep learning models if you want to use NodeJS as your web framework. Includes an extremely basic front-end built with Pug.

By default, this boilerplate runs a pre-trained resnet34 model that classifies Black Bears, Grizzly Bears, and Teddy Bears (h/t to fast.ai).

Extended from Node-Python by navjotts

Setup

In the root folder:

$ npm install
$ pip install -r requirements.txt

Demo with Bear Classifier model

$ npm start

Deploying your Image Classification Model

Export your Pytorch model as model-name.pth.

Put your .pth in /models.

At the top of model_fastai.py you'll see:

# SETUP HERE
YOUR_CLASSES_HERE = ['black', 'grizzly', 'teddys'] # Your class labels
NAME_OF_PTH_FILE = 'stage-2' # Name of your exported `.pth` file

1. Add your class labels
2. Add the name of your .pth
3. If you put your .pth file somewhere other than /models, replace PATH_TO_MODELS_DIR with that directory.
4. npm start

Something broken? https://twitter.com/zachcaceres