Fair Apparent Age Estimation

Using Convolutional Neural Network to create apparent age estimation model. Experiment with a number of methods to research their impact on the training process and different kid of biases.

ApparentAgeEstimation.pdf

The final report with the summary of contribution and the results discussion.

base_model.ipynb

Load the dataset, and the ResNet50 model. Use transfer learning to adjust the model for the given task and perform fine-tuning.

bias_simple.ipynb

Investigate the dataset with respect to the underlying biases. Implement simple methods like data augmentation and weighted samples to improve the training.

bias_custom.ipynb

Experiment with complex custom solutions for reducing biases, like complex weights or custom loss functions.