Face recognition app is developed to train and test the AI model on face recognition. Dataset used to train the model is LFW Dataset, while any other dataset can be used with similar format. Performance analysis is done using pairs.txt on the site.
Node.js as a backend tool and NPM Face Recognition as a face recognition library.
After training the model using approximately 13000 images of 6000 people, model performance is tested. On 6000 different cases, of different images of same person (expected a "match") or different people (expected a "mismatch"), models performance result is %98.62 success.