yuanzhaoYZ / face_detection_alignment

Face Detection and Alignment

face_detection_alignment

Face Detection and Alignment

Pre-requisites

Linux

conda create -n py27_tf012 python=2.7
source activate py27_tf012
pip install jupyter
pip install --upgrade ipykernel
pip install scipy
conda install PIL
source activate py27_tf012
conda install -c menpo opencv3
conda install -c menpo dlib

MAC

conda create -n py27_tf012 python=2.7
source activate py27_tf012
pip install jupyter
pip install --upgrade ipykernel
pip install scipy
conda install PIL
source activate py27_tf012
conda install -c menpo opencv3

install dlib on Mac Clean up dlib installed via conda:

conda uninstall dlib

Install cmake and boost-bython with homebrew:

brew install cmake
brew install boost-python

Build dlib from source:

git clone https://github.com/davisking/dlib.git
cd dlib/
mkdir build
cd build/
cmake .. -DDLIB_USE_CUDA=0 -DUSE_AVX_INSTRUCTIONS=1; cmake --build .
cd ..
source activate py27_tf012
python setup.py install --yes USE_AVX_INSTRUCTIONS --no DLIB_USE_CUDA

run preprocess.py

~/anaconda/envs/py27_tf012/bin/python preprocess.py \
--input-dir /Users/yuan/Downloads/workspace/out_aws_ibm_keras_age_not0_not2017_up_to_13248.csv_curated/inspection_not_needed_cropped_v2 \
--output-dir /Users/yuan/Downloads/workspace/out_aws_ibm_keras_age_not0_not2017_up_to_13248.csv_curated/tmp \
--crop-dim 512

Reference

• https://hackernoon.com/building-a-facial-recognition-pipeline-with-deep-learning-in-tensorflow-66e7645015b8