Improved OASIS UNet - s4802308

[olivermccarthy-uq]

Improved OASIS UNet - s4802308

[wangzhaomxy]

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

File Organizing: Please reorganise your files as following:

• Please move the "/script" folder to your project folder at "/PatternAnalysis-2025/recognition/OASIS-ImprovedUNet" instead of the current root directory "/PatternAnalysis-2025/recognition/".
• In addition, please restore the original README file in the folder /PatternAnalysis-2025/recognition/.
• Please also rename your custom folder to include both your project name and your student ID, as this directory will be merged with folders from other students.
• Kindly make these corrections; otherwise, the merge request will be rejected.

Problem Solving:

• The algorithm solves the problem appropriately.
• Accuracy in testing dataset (Dice): 0.632. Less than the benchmark.

Model and functions:

• It correctly uses PyTorch to construct the improved UNet 2D models and functions.
• NO data augmentation.
• Properly use the train/validation/test datasets.
• The Cross-Entropy loss function choice is inappropriate for a segmentation task. Dice loss should be the primary loss—either used on its own or combined with Cross-Entropy loss. Using only Cross-Entropy loss is not suitable in this context.

Code design: Good.

Code comment and docstring:

• Good code comments
• NO function docstrings
• NO header block

Difficulty: Easy. Using OAISIS dataset which is very simple to preprocessing, so it's Easy.

Additional Comments:

• Good commits.
• Good ReadMe design and report content. However, as a report, a conclusion section is expected to properly summarize your project. In addition, there are some display issues in the Readme, for example, the project structure.
• Using the load_oasis_data() function together with the OasisDataset class is not ideal. The load_oasis_data() function loads all images and masks into memory, which data scientists typically avoid. Consider using a Dataset and DataLoader to handle the data more efficiently. This is just advice—no marks will be deducted.

[gayanku]

Marking

Good/OK/Fair Practice (Design/Commenting, TF/Torch Usage)
	No design and implementation. Use DICE.	-2
	Spacing and comments.
	No Header blocks.	-1
Recognition Problem
	OK solution to problem. Acc low	-2
	Driver Script present.
	File structure present.
	Good Usage & Demo & Visualisation & Data usage.
	Module present.
	Commenting missing.	-1
	No Data leakage found.
	Difficulty : Easy. Easy. ImprovedUnet2D (OAISIS dataset which is very simpl)	-10
Commit Log
	Good Meaningful commit messages.
	Good Progressive commits.
Documentation
	Readme :Acceptable.	-1.5
	Model/technical explanation :Acceptable.	-2
	Description and Comments :Good.
	PDF submitted. Markup broken (Project Structure).	-2
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. Restructure + remove cache files for merge.	-2
	Request Description is missing.	-2
TOTAL		-25.5

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

[olivermccarthy-uq]

Hi @gayanku I have made the changes requested by you and Shakes. Hope they are okay. Thank you,