Face-Occlusion-Detect

A simple CNN face occlusion detect implemented with Tensorflow Keras. Detailed design doc, please reference to:

Depencies

dlib >= 19.17.0
tensorflow >= 1.12.0
keras >= 2.2.4
numpy >= 1.11.1
scipy >= 0.14
opencv-python >= 3.4.3

Usage

1. Data and trained models download

download link: https://pan.baidu.com/s/10LvoXEUGMTZjufd7R8jh4A
code: 0p5j
Download Cofw dataset and pretrained models in source code directory

2. Train

(1) prepare data

    python prepare_data_cofw.py --data_dir 'cofw data directory"

(2) train

python main.py --proj_dir \
           --proj_dir "./" \   #Project directory
           --input_size 96 \   #Input image size to train
           --batch_size 100 \  #train batch size
           --fine_tune False \ #Finetune VGG16 or not
           --epochs 100 \      #Train epochs
           --train True\     #Train or test

3. Test

(1) test on test_data

python main.py --proj_dir \
           --proj_dir "./" \   #Project directory
           --input_size 96 \   #Input image size to train
           --fine_tune False \ #Finetune VGG16 or not
           --train False\     #Train or test

(2) test on camera video stream data Need a camera device

python main.py --proj_dir \
           --proj_dir "./" \   # Project directory
           --input_size 96 \   # Input image size to train
           --fine_tune False \ # Finetune VGG16 or not
           --camera_test True