SynthEnsemble: A Fusion of CNN, Vision Transformer, and Hybrid Models for Multi-Label Chest X-Ray Classification

Published in 2023 26th International Conference on Computer and Information Technology (ICCIT)

Publication Link

https://ieeexplore.ieee.org/abstract/document/10441433

Training

The training folder contains the notebook links for all 6 models: ConvNextV2, SwinV2, MaxVit, VOLO D2, CoAtNet, DenseNet121.

Testing

The testing folder contains the notebooks for :

1. running trained models on test dataset with average weighted ensemble
2. finding optimal set of weights for weighted ensembling using differential evolution

TODO:

• Organize and upload notebooks used to train classical models as meta classifier