TF-FaceLandmarkDetection
Face landmark detection using tensorflow
Reproduction of the paper Deep Convolutional Network Cascade for Facial Point Detection
Requirements
- Python 3.4
- Tensorflow 0.10.0
Usage
- git clone https://github.com/mariolew/TF-FaceLandmarkDetection/edit/master
- Prepare data: You should have a text file, each line of the text file should have the format: image_path bbx_left bbx_right bbx_top bbx_bottom landmark1_x landmark1_y ... landmarki_x landmarki_y
- Modify the text file path and the path to store augmented images in augment.py and do python3 augment.py
- Modify some paths and params in model_train.py and do python3 model_train.py to train a face alignment model
- Modify some paths and params in model_eval.py and do python3 model_eval.py to evaluate the trained model
Note
This repo is based on https://github.com/luoyetx/deep-landmark and https://github.com/pkmital/CADL and is still ongoing.
Achievements
Level1: Done
Level2: TODO
Level3: TODO
References
[1] Deep Convolutional Network Cascade for Facial Point Detection
[2] deep-landmark
[3] Creative Applications of Deep Learning w/ Tensorflow