greenchacha / face_align_mtcnn

updated MTCNN based on ubuntu16.04

Welcome to FaceNote Pages

This Project focus on computer vision and artificial intelligence for face searching.

• opencv 3.4.2
• MTCNN
• FaceNet
• python 3.6.9
• tensorflow-gpu 1.2.1
• scipy 1.3.1
• imageio 2.6.1

The project has below components

1. Face detect and search validation

FaceNote

Install Requirments

pip3 install -r requirments.txt

Download Training Data

python3 facescrub_download.py
wget http://vis-www.cs.umass.edu/lfw/lfw.tgz

Align Images

python3 align_dataset_mtcnn.py ~/datasets/lfw/ ~/datasets/lfw_mtcnnalign_160 --image_size 160 --margin 32
python3 align_dataset_mtcnn.py ~/datasets/facescrub/ ~/datasets/facescrub_mtcnnpy_182 --image_size 182 --margin 44

Training classifier

python3 facenet_train_classifier.py --logs_base_dir ~/logs/facenet/ --models_base_dir ~/models/facenet/ --data_dir ~/datasets/facescrub_mtcnnpy_182 --image_size 160 --model_def inception_resnet_v1 --lfw_dir  ~/datasets/lfw_mtcnnalign_160 --optimizer RMSPROP --learning_rate -1 --max_nrof_epochs 200 --keep_probability 0.8 --random_crop --random_flip --learning_rate_schedule_file ../data/learning_rate_schedule_classifier.txt --weight_decay 5e-5 --center_loss_factor 1e-4 --center_loss_alfa 0.9 --batch_size 70 --epoch_size 100

If you have a pre-trained model, please add the --pretrained_model in the above command.

--pretrained_model $pretrained_model_name

For example:

~/tmp/models/facenet/20170223-125024/model-20170223-125024.ckpt-0

Raspberry

sudo mount -t cifs -o username=Minerva,password=sesame,vers=1.0 //192.168.8.1/sda2 /mnt/net_disk
sudo mount -o uid=pi,gid=pi /dev/sda1 /mnt/usb_flash

Having trouble with OpenCV? Having trouble with SciPy?

Face_align_MTCNN

Face_align_MTCNN