Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Feb 2, 2025

• Updated specCnn model to include batch normalisation

  jones139

  committed
  b5f2751

  Copy full SHA for b5f2751

Commits on Feb 1, 2025

• Added seiure durations for OSDB V1.8

  jones139

  committed
  5dac322

  Copy full SHA for 5dac322

Commits on Jan 25, 2025

• Added some event based True Positive Rate and True Negative Rate statistics calculations to testRunner

  jones139

  committed
  a1b931b

  Copy full SHA for a1b931b

Commits on Jan 24, 2025

• Removed debugging 'exit' statement from testRunner that was stopping it early

  jones139

  committed
  f2b9fc4

  Copy full SHA for f2b9fc4

• Added amber code to testRunner folder

  jones139

  committed
  7129142

  Copy full SHA for 7129142

• removed some debugging output and configured to test AMBER model

  jones139

  committed
  116c18f

  Copy full SHA for 116c18f

Commits on Jan 23, 2025

• Updated nnConfig.json for the actual configuration I used to train the amber model

  jones139

  committed
  24568ca

  Copy full SHA for 24568ca

Commits on Jan 20, 2025

• got AMBER model working!

  jones139

  committed
  a29cb92

  Copy full SHA for a29cb92

• Started adding Jamie Pordoy's AMBER model to nnTraining2 (not working yet)

  jones139

  committed
  e5d6352

  Copy full SHA for e5d6352

Commits on Jan 18, 2025

• Added spectrogram based CNN algorithm specAlg.py - showing very good seizure detection, but high false alarm rate (worse than OSDAlg, so need to look at why)

  jones139

  committed
  cd70a57

  Copy full SHA for cd70a57

• Added testRunner amberAlg ready to use with Jamie's AMBER model

  jones139

  committed
  cc672e0

  Copy full SHA for cc672e0

Commits on Jan 15, 2025

• Changed specCnn Model structure to try to make it less prone to over-training. Modified test script so it uses a balanced positive/negative test data set so it is quicker to run.

  jones139

  committed
  ad3f9dd

  Copy full SHA for ad3f9dd

Commits on Jan 12, 2025

• Fixed cnnModel and added configuration required for SpecCnnModel to nnConfig.json

  jones139

  committed
  2d9c2f9

  Copy full SHA for 2d9c2f9

• Converted values to g rather than milli-g so we work with smaller values, which should be better for the cnn

  jones139

  committed
  fecd0f1

  Copy full SHA for fecd0f1

• Got training and testing to run - had to adjust input shape because conv1d expects 3d data....

  jones139

  committed
  7aac039

  Copy full SHA for 7aac039

• Progressing getting the specCnnModel working (might have accidentally broken cnnModel though as I have introduced a configuration parameter to the constructor)

  jones139

  committed
  42180ff

  Copy full SHA for 42180ff

• Changed spetrogram resolution to make it more readable

  jones139

  committed
  bd525d2

  Copy full SHA for bd525d2

• I might have fixed the calculation of expected image shape based on step size...

  jones139

  committed
  b417ec1

  Copy full SHA for b417ec1

• Corrected synthesised data

  jones139

  committed
  ae0147a

  Copy full SHA for ae0147a

• Removed DC component by subtracting average for each 'slice' through spectrogram (so we can retain bin 0 for low frequency signals)

  jones139

  committed
  fe13ce9

  Copy full SHA for fe13ce9

Commits on Jan 11, 2025

• I spectrogram generator seems to be working in specCnnModel - just need to link this into CNN model now

  jones139

  committed
  3726774

  Copy full SHA for 3726774

• Started adding a spectrogram based CNN model - not working yet!

  jones139

  committed
  5fc4836

  Copy full SHA for 5fc4836

• Adjusting debug output

  jones139

  committed
  8953195

  Copy full SHA for 8953195

Commits on Dec 23, 2024

• Merge branch 'main' of https://github.com/OpenSeizureDetector/OpenSeizureDatabase

  jones139

  committed
  2b47c26

  Copy full SHA for 2b47c26

• Update curator tools readme

  jones139

  committed
  f1c114d

  Copy full SHA for f1c114d

• Added option to normalise accelerometer data and to apply a standard deviation threshold to testrunner data to avoid analysing very low movement datapoints.

  jones139

  committed
  2616dda

  Copy full SHA for 2616dda

• Added manually calculated stats to try to work out what the scikit ones are doing.

  jones139

  committed
  45cf407

  Copy full SHA for 45cf407

• Corrected confusion matrix calculation order so the axes are correct

  jones139

  committed
  b3ecc2a

  Copy full SHA for b3ecc2a

• Modified flattenData.py to only include datapoints within specified seizureTimes time range, and movement above the specified standard deviation threshold.

  jones139

  committed
  e8fcd2b

  Copy full SHA for e8fcd2b

Commits on Dec 22, 2024

• Removed 'Manual Alarm' (alarmState=5) from summary output and used highest automatically detected alarm level instead

  jones139

  committed
  4f27b21

  Copy full SHA for 4f27b21

• Removed some debugging output from makeOsdDb.py

  jones139

  committed
  0f9bc10

  Copy full SHA for 0f9bc10

Commits on Dec 20, 2024

• Updated requirements.txt for the new dataVisuliser examples - needed PyQt6 to run on Ubuntu 24.04

  jones139

  committed
  2f83c98

  Copy full SHA for 2f83c98

Commits on Dec 18, 2024

• Merge pull request #16 from jpordoy/data_visualiser

  Show description for 972aa88
  jones139

  authored
  972aa88

  Copy full SHA for 972aa88

• Update fft_plotter.ipynb

  jpordoy

  authored
  c72cace

  Copy full SHA for c72cace

Commits on Dec 17, 2024

• Distance fixed in IPYNB dataloader

  jpordoy

  committed
  a348bb3

  Copy full SHA for a348bb3

Pagination

Previous

Next