Project on Face Recognition using different types on CNNs, using the lfw dataset, was pretty fun. This was a project for course UE20CS302 on Machine Intelligence at PES, University.
Execution Steps:
- Import all the essential libraries
- Load the dataset from kaggle the link for this is : https://www.kaggle.com/datasets/jessicali9530/lfw-dataset
- Pre-Processing Involves cropping out the middle part and converting to greyscale (done for one iteration of code, rest are left colored)
- Run the model blocks and compile. It is defined as model(), multi_classifier() or resnet(), depending on the model
- Run the model for the dataset over the number of epochs specified (this will take a while)
- Evaluate for test data
- Run the accuracy metrics calculations, either confusion matrix, or accuracy score, formulae in code
- Import random file and predict with model.
- All these steps could also be achieved by downloading the .ipynb file and clicking run all on a jupyter notebook environment.