Implement Siamese Network for ISIC 2020 Skin Lesion Classification – s4778251

[lgyts]

Hi team,

This pull request contains my solution to the Siamese classification problem (Hard Difficulty Task 9)
After training, the model achieves an average test accuracy of around 81%, with balanced precision and recall (≈ 0.80).

README.md – Comprehensive documentation of the entire project including dataset, preprocessing, training procedure, results, and discussion.

dataset.py – Handles the ISIC 2020 dataset: metadata loading, patient-grouped splitting, augmentation, and data loaders.

modules.py – Contains model architectures: SiameseEncoder (ResNet-50) and BinaryClassifier (4-layer MLP).

utils.py – Helper utilities for plotting, feature extraction, saving sample images, and ensuring reproducibility.

params.py – Centralized configuration for all hyperparameters, paths, and augmentations.

train.py – Trains both the Siamese encoder (Triplet Margin Loss) and classifier (CrossEntropyLoss).

predict.py – Loads trained models to evaluate the test set and generate confusion matrix and metrics.

.gitignore – Excludes generated weights and datasets for clean repository submission.

/images/ – Contains figures used in the README (loss curves, confusion matrix, input samples).

[lgyts]

Please ignore the README uploaded to Gradescope — that version was incomplete and has been abandoned.

[hanemma7moud]

This is an initial inspection, no action is required at this point

Well Done!
Recognition Problem : total : 19.5

1. Solves problem: The solution is appropriate for the problem (4.5)

• Strong attempt, clear Siamese + classifier pipeline, proper explanation of triplet loss, classification stage, evaluation, and justification of augmentation + splitting.
  Minor issue (-0.5):
  Missing an embeddings visualization (t-SNE / UMAP).

2. Implementation functions : The code appears to be functional (3)

Good design: design is fairly strong(1)
Commenting: good(1)
Difficulty: Hard (10)

Great work , Thanks

[gayanku]

Marking

Good/OK/Fair Practice (Design/Commenting, TF/Torch Usage)
	Adequate design and implementation.	-1
	Spacing and comments.
	Header blocks.
Recognition Problem
	OK solution to problem.	-1
	Driver Script present.
	File structure present.
	Good Usage & Demo & Visualisation & Data usage.
	Module present.
	Commenting present.
	No Data leakage found.
	Difficulty : Hard. Hard Difficulty : Siamese
Commit Log
	Good Meaningful commit messages.
	Good Progressive commits.
Documentation
	Readme :Acceptable.	-1
	Model/technical explanation :Good.
	Description and Comments :Good.
	Markdown used and PDF submitted.
Pull Request
	Successful Pull Request (Working Algorithm Delivered on Time in Correct Branch).
	No Feedback required.
	Request Description is good.
TOTAL		-3

Marked as per the due date and changes after which aren't necessarily allowed to contribute to grade for fairness.
Subject to approval from Shakes