Dataset - http://vis-www.cs.umass.edu/lfw/
Youtube Video Reffered for help Video By Nicolas
- I have used used Docker for the hassle free setup of the tensorflow gpu with jupyter notebook
- All the training, preprocessing and inferences are done with this.
data/
├── anchor/
├── positive/
└── negative/
application_data/
├── input_image/
└── verification_images/
docker compose up
-
You can run the notebook in that environment but to add the positive and anchor images you have to use
imageCollector,py
file on your local machine due to the fact that docker doesn't connect to hardware directly. If you are a linux user you can add "/dev/video0:/dev/video0" in the devices in docker compose file. Otherwise its much of a hassle to set it up. -
Download the data,unzip it and move the images to desired folders using the notebook.
-
Go through the notebook for all other steps
Run
verification.py
I have configured it to run on CPU, you can change to GPU as well.