this facial verification app made for both tensorflow and pytorch
the model we will train need positive images and negative images and anchor images so we will collect these data
- first anchors and positive we will use cv2 to access our webcam and capture some images so the webcam_images.py will capture positive if you pressed 'p' in your keyboard and anchor if you pressed 'a'
python scripts/webcam_images.py - the negative example you can use any random images but we will scrap some from the web so the get_negatives.py script will scrab some images and save them localy and you can change the url to the site you want to scrab
python scripts/get_negatives.py - filter the negatives you will need to filter some of the unsuitalble images in the negatives with the filter negatives.py script
python scripts/filter_images.py- augment the data run the augment.py script to augment the data you collected
python scripts/augment.py- train the model
the data is ready we will start to train the model
tensorflow => you will open the tensorflow_model.ipynb notebook and run the cells to train the model and save it
pytorch => you will open the pytorch_model.ipynb notebook and run the cells to train the model and save it
- finally you will some inference
you will use cv2 to do some inference
tensorflow => you can run the tensorflow_inference script
python scripts/tensorflow_inference.py pytorch => you can run the pytorch_inference script
python scripts/pytorch_inference.py