Ensembles of GANs as a Data Augmentation Technique for Alzheimer research
summary
Uses ensembles of GANs to augment scarce Alzheimer MRI data, training DCGAN-style models and measuring quality with FID score (PyTorch). Compares three ensemble methods: snapshot/self-ensemble (best in experiments), naive ensemble of independently trained GANs, and a cascade approach that fails to help. Builds reproducible notebooks and generates synthetic 2D MRIs from an Alzheimer dataset; generated images improve diversity and can train classifiers that rival or sometimes beat classifiers trained only on real scans.
AI-written summary · about
chapters (17)
- 0:00Intro
- 3:05About Raul
- 4:59What is Alzheimer
- 8:03Why I started this project
- 13:37Deep learning
- 16:59Classification
- 21:12How it works
- 21:41Ensemble of GANs
- 22:54Three Techniques
- 23:12Self Ensemble
- 25:07Cascade
- 27:02Base Model
- 27:37Noise
- 27:47Gray Matter
- 28:19Seagun
- 28:56Conclusions
- 29:55Resources
related talks
30mVisual data abundant relevant labelled cheap Pick two
37mHarnessing the power of Generative Adversarial Networks (GANS) for supervised learning
40mVisual concept learning from few images
40mOverview of Generative Adversarial Networks (GANs)
01h 00mTutorial on Deep Learning IV
43mHitchhiker’s Guide to Generative Adversarial Networks (GANs)
13mNavigating Generative AI and Its Risks
25mEnsemble-X: Your personal strataGEM to build Ensembled Deep Learning Models for Medical Imaging