Code

Topic recognition - s4647936#165

Code

Open

tranvicky wants to merge 51 commits intoshakes76:mainshakes76/PatternAnalysis-2023:mainfrom tranvicky:topic-recognitiontranvicky/PatternAnalysis-2023:topic-recognitionCopy head branch name to clipboard

+876

Commits

Commits on Sep 17, 2023

• Added recognition branch and README for info.

  shakes76

  committed
  c3aff8b

Commits on Oct 4, 2023

• Initial setup for Siamese network Alzheimer’s disease classification.

  tranvicky

  committed
  7e5bc0e

• Renamed README.MD to README.md

  tranvicky

  committed
  8d8b7f2

• Update to contain rough outline for README.md

  tranvicky

  authored
  9aabb12

Commits on Oct 10, 2023

• Initial draft of Siamese network architecture in modules.py

  tranvicky

  committed
  7e03b36

• Added section in code to help determine dataset image dimensions

  tranvicky

  committed
  7d27279

• Merge branch 'topic-recognition' of https://github.com/tranvicky/PatternAnalysis-2023 into topic-recognition

  tranvicky

  committed
  7cacc40

• •TripletLoss setup: Create an initial structure for the triplet loss function in modules.py.

  tranvicky

  committed
  dea52f6

• Added TripletDataset to dataset.py to generate anchor, positive (same patient), and negative (different class) image triplets for Siamese training.

  tranvicky

  committed
  cbb5bea

Commits on Oct 14, 2023

• Refactored TripletDataset for enhanced adaptability, ensuring balanced anchor selection and optimized triplet formation for Siamese network training.

  tranvicky

  committed
  e096a1e

• Added image transformations and initialized train and test datasets in train.py

  tranvicky

  committed
  0628f8c

• Added functionality to display and save sample triplet images while checking for NaN and Inf values in the dataset

  tranvicky

  committed
  3387cc5

• Refactored TripletDataset to ensure patient-wise split for train-test data, improved triplet selection using patient IDs, and streamlined path extraction using os.path.basename

  tranvicky

  committed
  0d39d54

• Added testing to determine size of training sets after splitting

  tranvicky

  committed
  33b1d47

• Merge branch 'topic-recognition' of https://github.com/tranvicky/PatternAnalysis-2023 into topic-recognition

  tranvicky

  committed
  e5b4344

• Added additional data augmentation steps (random rotation, horizontal flip, and vertical flip) when in train mode

  tranvicky

  committed
  b04d801

Commits on Oct 23, 2023

• Integrated the training loop for the Siamese Network into train.py. Added visualization of sample images post-training.

  tranvicky

  committed
  cc2038f

• Fixed method save_image in train.y so that it properly saves images.

  tranvicky

  committed
  a1058d9

• Added validation/testing loop to evaluate model on test set post-training

  tranvicky

  committed
  22adb4c

• Implement predict.py to load the trained Siamese Network model and perform predictions on new sample data.

  tranvicky

  committed
  2896300

• Integrated PCA-based visualization for model embeddings in train.py and added functionality to save the 2D scatter plot as embeddings_pca.png.

  tranvicky

  committed
  561d8a3

• Changed embedding plots from PCA to T-SNE

  tranvicky

  committed
  3c30104

• Added functionality to extract and store embeddings for the entire dataset in train.py This will enable subsequent classifier training on Siamese network embeddings.

  tranvicky

  committed
  eda84c8

• Modified dataset.py to include labels, allowing for easier extraction and use of labels when processing embeddings post Siamese network training.

  tranvicky

  committed
  0cc7e8d

• Added SimpleClassifier class in the neural network module to facilitate the classification of Siamese network embeddings into AD or NC categories

  tranvicky

  committed
  6a77f09

• Implemented the training loop for the Simple Classifier using extracted embeddings from the Siamese Network. Utilized the Adam optimizer and CrossEntropyLoss for training.

  tranvicky

  committed
  3e7fc30

• Added evaluation phase for the Simple Classifier using the test embeddings. The accuracy metric is computed to assess the classifier's performance on the test set.

  tranvicky

  committed
  f0ba8da

• Changed labels to detect if AD or NC in it's path.

  tranvicky

  committed
  5f9c135

• Added method plot_confusion_matrix to plot confusion matrix so can visualise.

  tranvicky

  committed
  2afb4e3

• Changed train.py so code can print and out plot of train vs validation losses.

  tranvicky

  committed
  3010854

• Added tracking and visualization for the classifier's training and validation losses. This enhancement aids in monitoring the classifier's performance across epochs.

  tranvicky

  committed
  879bc08

• Changed code to fix issue of tsne graph displaying only one colour

  tranvicky

  committed
  5f66d81

• Add code for SNN to be stopped early if validation loss doesn't improve for 5 consecutive epochs).

  tranvicky

  committed
  6f2924f

Commits on Oct 24, 2023

• Added changed to train.py to resolve issue of incorrect indexing within classification evaluation.

  tranvicky

  committed
  c4a4810

• Changed train.py code to save plot of siamese training vs losses.

  tranvicky

  committed
  2b5d722

Commits on Oct 25, 2023

• Solved issues with pconfusion matrix plot not showing and issue with incorrect embedding sizing.

  tranvicky

  committed
  a9e5b91

• Introduced more convolutional layers and added dropout after each fc layer

  tranvicky

  committed
  bc90ea3

• Adjusted hyperparameters for learning rate and batch size.

  tranvicky

  committed
  a11b716

• Updated README.md to fill out all of the uncompleted sections.

  tranvicky

  committed
  ca94001

• Changed code to resolve dimension bugs after changed number of fc layers.

  tranvicky

  committed
  1af6bb3

• Updated README.d to include visualisations/plots produced during training and also added results summary.

  tranvicky

  committed
  dbdaff3

• Updated README.md to reflect updated dependencies.

  tranvicky

  committed
  28c03b8

• Updated predict.py so that if user wants to visualise the model, they can now just run the file. Updated with saving all visualising plots.

  tranvicky

  committed
  bad03a6

• Updated README.md to detail a more in-depth setup guide and running for users.

  tranvicky

  committed
  bf8a1a4

• Updated train.py to resolve issue about numpy to tensor error.

  tranvicky

  committed
  53f5e2f

• Updated references for README.md

  tranvicky

  committed
  1fb431b

• Removed unecessary print statements from train.py, modules.py and dataset.py.

  tranvicky

  committed
  5814d1b

• Provide folder for images/plots/visualisations needed for README.md.

  tranvicky

  committed
  8d8ad25

• Update README.md to resolve issue of images only showing links instead of images and to display images properly.

  tranvicky

  authored
  77b19d5

• Updated README.md and fixed image pathways.

  tranvicky

  authored
  fb57496

• Changed variable outputs to resolve tensor error in train.py.

  tranvicky

  committed
  6aee14a