This is completly based on deep learning nueral network and implented using Tensorflow framework. Here you will get how to implement fastly and you can find code at github and uses is demonstrated at YouTube.
- Tensorflow (1.10.0)
- Scipy (1.1.0)
- Scikit-learn (0.19.1)
- Opencv (3.4.4.19)
For ipmlementation and run this code follow this BLOG link. Implemention in video is shown YouTube.
Download FaceNet Pre-trained models: 20170512-110547.pb - [Google Drive] / [Baidu Drive]
Training using the VGGFace2 dataset - [Link]
-
Put your Pre-trained models 20170512-110547.pb file under './model/' folder.
-
data_preprocess.py put img under './train_img' folder and run data_preprocess.py output preprocessed img to './pre_img' folder.
-
train_classifier.py training your own data under './pre_img' folder and output classifier.pkl file to './classifier/' folder.
-
identify_face_image.py face recognition on your own images(change 'input_image' path)
-
identify_face_video.py face recognition on your own videos(change 'input_video' path)
Hopenet is an accurate and easy to use head pose estimation network. Models have been trained on the 300W-LP dataset and have been tested on real data with good qualitative performance.
For details about the method and quantitative results please check the paper.
new Conan-Cruise-Car example video
To use please install PyTorch and OpenCV (for video) - I believe that's all you need apart from usual libraries such as numpy. You need a GPU to run Hopenet (for now).
To test on a video using dlib face detections (DLIB_MODEL: mmod_human_face_detector.dat):
python3 code/test_on_video_dlib.py --snapshot PATH_OF_PRE-TRAINED-MODELS --face_model PATH_OF_DLIB_MODEL --video PATH_OF_VIDEO --output_string STRING_TO_APPEND_TO_OUTPUT --n_frames N_OF_FRAMES_TO_PROCESS --fps FPS_OF_SOURCE_VIDEO
To test on a video using your own face detections (we recommend using dockerface, center of head will be smoother):
python3 code/test_on_video_dockerface.py --snapshot PATH_OF_SNAPSHOT --video PATH_OF_VIDEO --bboxes FACE_BOUNDING_BOX_ANNOTATIONS --output_string STRING_TO_APPEND_TO_OUTPUT --n_frames N_OF_FRAMES_TO_PROCESS --fps FPS_OF_SOURCE_VIDEO
Face bounding box annotations should be in Dockerface format (n_frame x_min y_min x_max y_max confidence).
300W-LP, alpha 1, robust to image quality
For more information on what alpha stands for please read the paper. First two models are for validating paper results, if used on real data we suggest using the last model as it is more robust to image quality and blur and gives good results on video.
Please keep in mind that testing instructions to reproduce the paper results will be added very soon.
This work is still in progress - we are obtaining better results and will also be updating this README with instructions. Please open an issue if you have an problem.
Some things that will be added:
- Test script for images
- Docker image
- Instructions for all scripts
- Better and better models
- Videos and example images!
If you find Hopenet useful in your research please consider citing:
FaceNet - "FaceNet: A Unified Embedding for Face Recognition and Clustering"
Face alignment using MTCNN - "Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks"
@AI Sangam
For details about the method and quantitative results please check the paper. Nataniel Ruiz, Eunji Chong, James M. Rehg Georgia Institute of Technology