Ensemble-X: Your personal strataGEM to build Ensembled Deep Learning Models for Medical Imaging
summary
Ensemble combines DenseNet, InceptionV3 and a custom CNN with non-linear weights and training that freezes all but the final dense layers (freeze all layers except FC) then retrains, using Adam optimizer for 250 epochs. Evaluations span OCT retinal, malaria cell and skin‑lesion datasets, report ~98.5% top accuracy (1.47% error) on the test set, low overfitting (about 1 in 15 runs), external unseen validation and per‑image confidence scores, and delivers grid visualisations and a planned mobile/web deployment.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 2:14Medical imaging history
- 3:20Ensembling in deep learning
- 5:07Medical imaging challenges
- 7:35Algorithm overview
- 9:42Phase one training
- 11:49Phase two ensemble
- 12:51Why ensembling works
- 14:57Confusion matrix analysis
- 15:39Overfitting analysis
- 16:20Results presentation
- 18:07Benchmark comparison
- 20:34Planned enhancements and challenges
- 23:20Closing remarks
related talks
31mEnsemble X Your personal strataGEM to build Ensemble
30mTransfer learning for image recognition in healthcare...
55mDeep Learning in Medical Image Diagnostics
56mTutorial on Deep Learning II
15mBrain Tumor Detection and Classification Using Deep Learning Models
20mMulti-algorithm Ensemble Learning at Scale: Software...
32mEnsembles of GANs as a Data Augmentation Technique for Alzheimer research
25mDeep Learning in Biomedicine: Challenges and Opportunities