Implements Siamese Neural Network for One Shot Image Recognition
- Unzip the file at this location and add those files in 20180402-114759 folder (These are pre-trained weights)
- Pre-trained models are alread present in mtcnn folder
- Python modules required are
- tensorflow==1.13.1
- numpy==1.16.3
- sklearn==0.20.0
- cv2==3.4.3
- flask==1.0.2
- werkzeug==0.14.1
- tqdm==4.28.1
- matplotlib==2.2.4
- pandas==0.23.4
Enter your images in the folder data Make sure you enter images class-wise, i.e. if you have photos of 100 different people, make 100 folders inside data and keep at least 1 image inside each of it.
Note: Seems like there is some issue with using .png images. Try using .jpeg or .jpg
Run the python files in the order
python create_face_embeddings
python align_dataset_mtcnn
python rest-server
The last commands launches a server at http://127.0.0.1:5000/ on which the FR system can be accessed
In case you want to train your own models, follow these 2 links:
- Facenet training: https://github.com/davidsandberg/facenet/wiki
- MTCNN training: https://github.com/wangbm/MTCNN-Tensorflow
Feel free to ping me in case anything is broken. Thanks for visiting. Cheers!