3D U-Net for Prostate MRI Segmentation – Janvhi Sharma (s4975045)

[JS-0911]

Summary

Implements a 3D U-Net for prostate MRI segmentation (binary). Trained on HipMRI Study Open.
Achieved validation Dice ≈ 0.80 with smooth loss convergence.

Files (as required)

• modules.py — UNet3D, conv blocks (documented)
• dataset.py — MRI loader, normalisation, alignment
• train.py — training/validation, saves loss_curve.png & dice_curve.png
• predict.py — CLI to load ckpt and save visual comparisons
• README.md — full documentation, usage, dependencies, data splits

Evidence

See visuals/: loss_curve.png, dice_curve.png, comparison_* PNGs.
No datasets/model weights committed. Reproducibility instructions included.

Difficulty

Hard — 3D UNet with volumetric training on HPC (A100).

Notes

• Binary task ⇒ single-channel sigmoid. One-hot not applicable.
• Ready to incorporate any feedback; happy to adjust paths/figs if requested.

[24msingh24]

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

2D UNet – Prostate MRI Segmentation → Easy

Category		Marks	Comments
Algorithm solves the problem	5	5	Correctly implemented a 3D Improved U-Net for prostate segmentation and the model achieves a Dice score of ~0.80 on validation.
Implementation functions as intended	3	3	Seems to be fully functional
Good design	1	1
Commenting	1	1
Algorithm above Normal Difficulty	5	5	Normal
Algorithm is Hard difficulty	5	0	Normal
Section IV : Max mark 10 from 20	15

• Discussion: None provided. This is needed.

Suggestions/Notes:

• <span style="box-sizing: border-box;”>

[gayanku]

Marking

Good/OK/Fair Practice (Design/Commenting, TF/Torch Usage)
	Good design and implementation.
	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 : Normal. Normal. Unet 3D	-5
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).
	Feedback action require: Feedback marks possible +2 if the requested changes are made. Files are in wrong location. Create a folder with student ID for these, for merge.	-2
	Request Description is good.
TOTAL		-9

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

[JS-0911]

Hi,
I’ve updated the file structure as requested and moved all project files into the correct student ID folder.