Face parsing via Interlinked Convolutional Neural Network(Pytorch reimplement)

Paper

Description

This is a pytorch implementation of Zhou et al (2015). NOTICE: We have released a upgraded version of iCNN naming STN-iCNN, check the paper or code for more information.

The network archtecture is as following:

Pretrained model

Stage1+Stage2

Prepare datasets

Stage1 Dataset

1. ![Download](http://pages.cs.wisc.edu/~lizhang/projects/face-parsing/SmithCVPR2013_dataset_resized.zip) Smith et al. Resized HelenDataset 
2. Unzip it into ./datas/helen/

Stage2 Dataset

1. python3 ./utils/extract_parts.py

Visual Test

Run Jupyter Notebook: visual_test.ipynb

How to run

clone project

git clone https://github.com/aod321/icnn-face

install requirements

cd icnn-face pip install -r requirements.txt

train stage1

python train_stage1.py

train stage2

python train_stage2.py

all the checkpoints can be found at checkpoints_{uuid} or checkpoints_{parts_name}_{uuid}

Results

Comparison with State-of-the-art Methods on HELEN

Others