MicroExpNet.js

Translation of MicroExpNet to JavaScript using TensorFlow.js.

Requirements

• Python 3.6.2
• tensorflow 1.8.0
• tensorflowjs 0.4.2
• Flask 1.0.2 (For serve model file)
• face.png (84x84x3)

Usage

1. Clone repositories

git clone https://github.com/walkingmask/microexpnetjs.git
cd microexpnetjs
git clone https://github.com/cuguilke/microexpnet.git
cp generator.py microexpnet/

2. Generate pb file

cd microexpnet
python generator.py
# this will makes ./output_graph.pb

3. tensorflowjs-converter

tensorflowjs_converter --input_format=tf_frozen_model --output_node_names='Add_1' ./output_graph.pb ../demoapp/static/js/model

4. Prepar face.png

cd ..  # microexpnetjs
mkdir demoapp/static/img
cp /path/to/face.png demoapp/static/img/
# size of face.png must be 84x84x3

5. Run demoapp

python demoapp/app.py
# You will open http://localhost:5000

Todo

• Add face detection
• As realtime
• GitHub Pages