Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Topic recognition #485

Open
wants to merge 16 commits into
base: topic-recognition
Choose a base branch
from

Conversation

thomas2001u
Copy link

@thomas2001u thomas2001u commented Nov 13, 2022

This is my second pull request, in my first request I accidentally did a request from the incorrect branch.
The initial request can be found here: #465
I didn't realise I made this mistake until today, I hope that's forgivable, thanks. Commits are dated before the due date.

This is the PatternFlow project submission for COMP3710 by Thomas Barton (s4641500). This includes a VQVAE with Pixel CNN trained and tested on the OASIS dataset. The algorithm trains using png slices of OASIS brains to create a seed for outputting slices based on an input gathered from encoding brains. Testing on this input can show a brain that looks either more healthy or unhealthy, based on the training dataset used. It can be used to gauge the health of a brain by comparing outputs with the VQVAE.

EDIT: Please also note that two README files have been committed. The one under XUE4645768 was automatically added for some reason. It is not mine so please disregard it.

Created modules.py and partially implemented vector quantizer class.
Added encoder and decoder for keras layers
Basic trainer class. Train steps which backpropagate and check loss.
Implemented data loading and preprocessing.
Moved VQVAE train class from train to modules.
Implement pixel cnn. Basic functionality works.
Added class methods in modules.py allowing models to be saved in .h5 format. Model fully trains but performance metrics are yet to be separated from train.py.
Put many lines of code into functions and made it overall neater.
Made some parameters clearer and neatened up.
Modified the VQ initialiser to allow loading the model from file. Predict.py now implements necessary parameters used to train the model for testing. Train.py has been shortened greatly.
Wrote readme.
Changed epoch number.
This reverts commit 8ce3cab.
@thomas2001u thomas2001u mentioned this pull request Nov 13, 2022
@shakes76
Copy link
Owner

Good Practice (Design/Commenting, TF/Torch Usage)

Adequate use and implementation (code seems to be there, but can't verify if working, no outputs or training loss) -5
Good spacing and comments
Header blocks missing -1

Recognition Problem

Solves problem (no results provided, no convergence plotted, no generations) -3
Driver Script present
File structure present
Shows Usage & Demo & Visualisation & Data usage (no plots or outputs) -2
Module present
Commenting, some missing (see train.py for example) -1
No Data leakage (can't verify generations or data sets used) -2
Difficulty: Hard

Commit Log

Meaningful commit messages, could be more descriptive -1
Progressive commits used, could not determine any commits before 21 Oct because of incorrect PRs -1

Documentation

ReadMe minimal (no architecture, background etc.) -2
Good Description and Comments
Markdown used PDF submitted

Pull Request

Successful Pull Request (Working Algorithm Delivered on Time in Correct Branch)
Feedback required, remove changes to other student files, resolve conflicts -2
Request Description good
LATE, but reason give is acceptable because original PR and commits were made in time

@thomas2001u
Copy link
Author

Tried reverting changes to XUE4645768 folder by reverting to previous commit then force pushing that to PR. It's appearing to delete other students' files though? This sort of thing wasn't covered in the required git courses so I'm having trouble considering I only had half the day to figure this out.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants