superdianuj / MRI_condensation

Dataset Condensation Via Distribution Matching of MRI dataset

MRI_condensation

In this repo, I perform dataset condensation via distrbution matching (https://github.com/VICO-UoE/DatasetCondensation), without differentiable siamese augmentation, on MRI classification dataset (https://www.kaggle.com/datasets/sartajbhuvaji/brain-tumor-classification-mri)

Conensation of Whole Unbalanced Dataset

Running Time (seconds)	Accuracy over Real Training	Accuracy over Real Testing	Condensed Images per Class
676.7407	43.0662	29.9492	10

Condensation of Balanced Dataset

Running Time (seconds)	Accuracy over Real Training	Accuracy over Real Testing	Condensed Images per Class
649.1408	25.1282	20.3046	10

Condensation of 20 MRI images/class into 2 images/class

Running Time (seconds)	Accuracy over Real Training	Accuracy over Real Testing	Condensed Images per Class
311.9385	52.5	27.4112	2