OASIS brain data set using VQVAE - Daniel Miller 45810536 - #458

Open

dapmiller wants to merge 30 commits into

shakes76:topic-recognitionfrom

dapmiller:topic-recognition

Commits on Oct 18, 2022

Commits on Oct 19, 2022

created a function to batch data before training in dataset.py
dapmiller
committed
9ee75d3

Commits on Oct 20, 2022

Successfully trained data. Safety Check Commit. Removed Batching function from dataset.py. Reformatted the functions in modules.py -> encoder and decoder are now in terms of modules instead of sequ…
dapmiller
committed
605cb17
Created functions in train.py that compute structural similiarity. It currently does not work however. Training for data is also working, however there is a high loss.
dapmiller
committed
1e99576

Commits on Oct 21, 2022

Edited the ssim in train.py. Also added references to dataset.py anhd modules.py
dapmiller
committed
d361777
Fixed bug that was stopping loss reduce in model.fit. Currently implementing results for post model training
dapmiller
committed
235585b
Create Images
dapmiller
authored
b0d498b
Delete Images
dapmiller
authored
9b90b9f
Edited train.py bug in training. It now runs and the losses appear to decrease realistically. The bug occurred due to the value of the variance variable being feed into model.fit.
dapmiller
committed
f396991
Fixed Bug in model.fit that caused loss. It was due to variance being calculated incorrectly. Wrote draft read.me file
dapmiller
committed
3aeadda
Fixed Bug in model.fit that caused loss. It was due to variance being calculated incorrectly.
dapmiller
committed
a8f82d0
Merge branch 'topic-recognition' of https://github.com/dapmiller/PatternFlow into topic-recognition
dapmiller
committed
92073ce
Fixed Bug in model.fit that caused loss. It was due to variance being calculated incorrectly.
dapmiller
committed
ab702fe
Trying to add images
dapmiller
committed
6899684
Added Graph
dapmiller
committed
63a9a18
Cleaned up Files and finsihed off REadme file. Commented out the pixelcnn stuff as ran out of time to finish it
dapmiller
committed
f5db768
Fixed ssim to read 0.74 not 74
dapmiller
committed
09450f8

Navigation Menu

OASIS brain data set using VQVAE - Daniel Miller 45810536 - #458

OASIS brain data set using VQVAE - Daniel Miller 45810536#458
dapmiller wants to merge 30 commits into
shakes76:topic-recognitionfrom
dapmiller:topic-recognition

Commits on Oct 11, 2022

Trying something meaningful

Created required files

Commits on Oct 14, 2022

Provided brief description for each file of what is to be included

Added 2 new functions into datasey.py which one downloads the oasis dataset and the other loads the training images into a numpy array

Commits on Oct 15, 2022

Add comments to function for explanation and understanding in dataset.py file for normalising and preprocessing training data

Added functions to dataset.py to load labels and also process them with one hot encoding. Train.py now calls this for training data.

Added calls to load validation and test labels in train.py from dataset.py

Added brief summary to Read.me file about model

Safety commit of working loading and processing of data before implementing VQVAE structure.

Created encoder, decoder and overall vq-vae functions in modules.py however, quantised layer function still needs to be implemented and is currently preventing the code to run.

Commits on Oct 18, 2022

Finished building model in modules.py. It compiles and prints out summary in train.py

Finished building model in modules.py and it successfully compiles in train.py

Testing commit

Commited the module.py file and removed unnecessary code

Commits on Oct 19, 2022

created a function to batch data before training in dataset.py

Commits on Oct 20, 2022

Successfully trained data. Safety Check Commit. Removed Batching function from dataset.py. Reformatted the functions in modules.py -> encoder and decoder are now in terms of modules instead of sequ…

Created functions in train.py that compute structural similiarity. It currently does not work however. Training for data is also working, however there is a high loss.

Commits on Oct 21, 2022

Edited the ssim in train.py. Also added references to dataset.py anhd modules.py

Fixed bug that was stopping loss reduce in model.fit. Currently implementing results for post model training

Create Images

Delete Images

Edited train.py bug in training. It now runs and the losses appear to decrease realistically. The bug occurred due to the value of the variance variable being feed into model.fit.

Fixed Bug in model.fit that caused loss. It was due to variance being calculated incorrectly. Wrote draft read.me file

Fixed Bug in model.fit that caused loss. It was due to variance being calculated incorrectly.

Merge branch 'topic-recognition' of https://github.com/dapmiller/PatternFlow into topic-recognition

Fixed Bug in model.fit that caused loss. It was due to variance being calculated incorrectly.

Trying to add images

Added Graph

Cleaned up Files and finsihed off REadme file. Commented out the pixelcnn stuff as ran out of time to finish it

Fixed ssim to read 0.74 not 74