Visualize Activation Maps to Understand CNN Filters for Age Detection

This project aims to predict the age of individuals from images using a pre-trained ResNet50 model. We preprocess the dataset, train the model, and visualize the results using Grad-CAM.

Project Details

• Dataset: UTKFace (https://www.kaggle.com/datasets/jangedoo/utkface-new)
• Model: ResNet50 Fine-Tune

Dynamic Stats

• Last Update: 12/06/2024
• Total Images Processed: 15382
• Test Loss: 10.94
• Test MAE: 8.41

Age Distribution Before Filtering

Age Distribution After Filtering

Training & Validation Loss

Grad-CAM Heatmaps

Google Drive Link

https://drive.google.com/drive/folders/1ZFa_j2d6DmOtD-uEYqXtbjVGQ-9CEgCw?usp=drive_link

Author: Apurv Chudasama

License

This project is licensed under the MIT License - see the LICENSE.md file for details.