Title | Face Recognition |
---|---|
Team | Dang Quoc Vu - Vu. qvdang.96@gmail.com |
Predicting | Build a model for Vietnamese football players face recognition. The data is collected from google and captured from videos. Need to build a model that predicts people in the large image. |
Data | capture images from videos (https://www.youtube.com), images from https://www.google.com/search. Link:https://drive.google.com/file/d/1P4Vf5wsrsWZc8AHH7jLbceiG5KS3580Y/view?usp=sharing |
Features |
|
Models |
|
Future | impove the speed and accuracy in detecting and recognizing the human faces |
References | [1] https://arxiv.org/pdf/1502.00046.pdf [2] https://medium.com/analytics-vidhya/face-recognition-with-vgg-face-in-keras-96e6bc1951d5 [3] https://www.pyimagesearch.com/2018/01/22/install-dlib-easy-complete-guide/ [4] https://sefiks.com/2018/08/06/deep-face-recognition-with-keras/ |
Results | Score: > 0.96 |
|cnn_model/
| |-- mmod_face_detector.dat
| |-- vgg_face_net_weight.h5
|dataset/
| |--data_images/
| |--video/
| |--(video)
| |--video_frames/
| |--(frames of video)
| |--test_image/
| |--(images)
| |--train_image/
| |--Football-player[1]/
| |--(images)
| |--Football-player[2]/
| |--(images)
| |--Football-player[3]/
| |--(images)
| |--Football-player[4]/
| |--(images)
|other_file/
| |--(processed data files)
|src/
| |--(files .py)
|test_result/
| |--(images)
|trained_model/
| |--(model file .h5)
Training:
python3 face_train.py --p <bool>
- True : Process data before training
- False : Use current processed data to train
Testing:
python3 face_test.py --image <image_name>
- Paste the test-image into /dataset/test_image/
- The result image is saved in test_result folder with the same name
Result:
The picture below shows the trained model is able to classify 11 football players in Vietnam national team successfully.