Diagnosis of Dermoscopic Images using Multi-Sizing Ensemble-Based Deep Learning Method

Link for the Presentation - https://docs.google.com/presentation/d/1FBZ0tmoJJjrAS50cO-lpZKxSRPU5a5Ka7sM7yRlz8rs/edit?usp=sharing

To View the Validation Results, Have a Look at the Presentation.

How to Run the Code?

1. Train the Architectures with your Preferences and/or Requirements, or better Devise your Own using the Referred PyTorch Example.
2. Play with the Hyperparameters and Different Architecture Configurations. It is my favourite part.
3. Have Fun Transfering/Learning the Learnings.
4. Dataset - https://doi.org/10.7910/DVN/DBW86T

Note : Comments are provided for easier explanation. For more information, contact me!

Challenge Results (MAIA - 3rd Semester)

Our Methodology won the 1st Place in both the Binary with 90.8% Accuracy and Multi-Class 92% Accuracy (0.857 Kappa Score) on the Test Dataset in the Deep Learning Based Classification Challenges.