This is the coursework for CM50265 Machine Learning 2 (Semester 2 - 2022) at the University of Bath.
Grade: 97/100
The goal of this project is to develop CNN models to predict the age and gender of people given face images input. The dataset is a subset of 5,000 images from the UTKFace dataset.
Two models are built, one from scratch, another with transfer learning. The final report can be found here.
The model has a shared feature extraction pipeline, and two separated branches for age and gender predictions.
| Number of Epochs | Age MAE | Gender Accuracy |
|---|---|---|
| 85 | 6.12 | 89.40% |
The model uses DenseNet pretrained on ImageNet for feature extraction, and separate branches of fully connected layers for age and gender prediction respectively.
| Number of Epochs | Age MAE | Gender Accuracy |
|---|---|---|
| 143 | 5.75 | 90.53% |