One sentence project description: Develop a machine learning model that can perform age detection.
Figure author: Tommy Zhou
Dataset: UTKFaces
Images are split into 8 imbalanced age classes. To decrease bias in prediction, we preprocessed the dataset so the 8 classes are balanced, each class contains 770 images.
Data Processing:
- Identify faces: Haar-cascade
- We developed a small algorithm that is able the choose the "main face" from a given image
Five convolutional layers from AlexNet + Five fully connected layers.
Final Accuracy: 61.27% validation accuracy
Carter Huang
Tommy Zhou
Yu Xin Cheng