Description

This was my very first attempt to train a model from scracth. I had a goal of analyzing two well-known face recognition datasets (VGG Face and UTKFace) and explore if there are biases in two datasets by training two identical models (ResNet) and compare the accuracy of each model. We added Tiny ImageNet for an artificial dataset.

My responsibilities were:

• Implementation of model training module in the main routine
• Utilizing Nvidia CUDA to speed up the training process
• Creating a test-set for verification of models' performance
• Model performance visualization

Skills