Tensorflowjs and Keras

setup

$ pip install tensorflowjs

convert Keras model to Tensorflowjs model with command

$ tensorflowjs_converter --input_format keras models/mobilenet_v2.h5 models4js/mobilenet_v2

convert Keras model to Tensorflowjs model

for VGG16
$ python keras2tensorflow_forVGG16.py

for MobileNet V2
$ python keras2tensorflow_formobilenet.py

In terms of mobilenet_v2, if you got an error, you would need to update Keras

$ pip install -U keras

model.json is a meta-data file.

How to try this

clone this repo
$ git clone https://github.com/sleepy-maker/tensorflowjs-and-keras.git

$ cd image_classification/app

$ npm install

$ node server.js

demo

image_uploader version https://sleepy-maker.github.io/tensorflowjs-and-keras/image_classification/image_uploader/app/public/ webcam version https://sleepy-maker.github.io/tensorflowjs-and-keras/image_classification/webcam/app/public/

ToDo

• Add a function which allows users to select a model(VGG16/MobileNet)
• Optimize model loading process
• Add webcam function
• Rewrite with reactjs with typescript
• Add new functions