Issues
- 3
Trying to get in touch regarding a security issue
#406 opened by zidingz - 0
Add data QC module
#370 opened by fazamani - 0
- 0
fix seeds in train.py and test.py to make results perfectly reproducible on a single GPU - enhances the robustness of results
#382 opened by maxkazmsft - 1
add unit test to check that each input slice has uniform distribution - facilitates uniform sampling of training data for better model accuracy
#319 opened by maxkazmsft - 0
debug test.py scoring of trained models - facilitates correct results when scoring trained models
#393 opened by maxkazmsft - 0
add GPU profile information to multi-GPU train and test scripts - facilitates ability to view possible code bottlenecks
#348 opened by maxkazmsft - 0
determine the best configuration parameters for multi-GPU training - facilitates better multi-GPU results
#383 opened by maxkazmsft - 0
README documentation fixes - facilitates clear understanding of the content of the repo
#381 opened by maxkazmsft - 0
update documentation with SEResNet and data requirements - facilitates better user onboarding experience
#354 opened by maxkazmsft - 0
- 1
Add docker tests - ensures docker image doesn't break while new features are added
#358 opened by yalaudah - 0
debug slow runtime in multi GPU training - figure out why multi-GPU training does not produce much of a speedup
#385 opened by maxkazmsft - 0
bring back Penobscot dataset and treat it as BYOD - adds another dataset which we can quote results on
#321 opened by maxkazmsft - 0
run Penobscot dataset through tensor format - facilitates uniform data loader and converter sample code
#351 opened by maxkazmsft - 0
add the learning rate to the tensorboard
#371 opened by fazamani - 0
investigate the reproducibility problem with patch_deconvnet for small training data set
#372 opened by fazamani - 0
Remove depth channel from Tensorboard visualization - makes grayscale images appear as such in TensorBoard logging
#355 opened by yalaudah - 1
enable multi-GPU training with Docker image - facilitates multi-GPU training on any Linux OS
#378 opened by maxkazmsft - 0
- 0
debug test.py patch padding before model scoring - facilitates scoring correctness and higher scoring accuracy
#330 opened by maxkazmsft - 0
- 0
follow up with legal regarding temporary hosting of datasets for builds - facilitates faster setup job runtime
#353 opened by maxkazmsft - 0
extend config object in train and test and pass it to data loaders - facilitates the use the config variables by lower level functions
#352 opened by maxkazmsft - 1
add correctness data and class size unit test metrics - facilitates certainty in correct data flow
#317 opened by maxkazmsft - 0
make tests simpler (reduce test data size) - facilitates faster builds for PRs
#346 opened by maxkazmsft - 0
provide pre-trained SEResNet model with high accuracy - facilitates faster onboarding with pre-trained SEResNet model
#365 opened by maxkazmsft - 0
change HRNet notebook to use SEResNet - removes possibility of model malfunction in the notebook
#350 opened by maxkazmsft - 0
move contributions from CSE team into main folders - facilitates more user friendly repository with code samples
#349 opened by maxkazmsft - 0
add test for multi-GPU train.py run - facilitates bug-free train.py for multi-GPU training
#367 opened by maxkazmsft - 0
add multi-GPU training to train.py Dutch F3 local - facilitates faster training times
#320 opened by maxkazmsft - 0
operator1 and operator2 close-out
#374 opened by maxkazmsft - 0
- 2
image normalization should be wrt the volume image - facilitates correctly displayed image colormap range
#342 opened by fazamani - 1
debug multiple classes showing up on binary synthetic data - facilitates correct mask visualizations representative of the number of classes
#326 opened by yalaudah - 1
debug checkerboard synthetic dataset having empty padded images as input - facilitates higher training accuracy and training correctness
#328 opened by maxkazmsft - 0
Azure ML training pipeline improvement comments - facilitates better user onboarding for the pipeline
#366 opened by maxkazmsft - 0
add pre-trained models to test section of the HRNet notebooks - facilitates faster onboarding with more pre-trained models
#364 opened by maxkazmsft - 0
- 0
make pre-run notebooks for segyconverter utility and SEResNet available in main README - facilitates faster quick-start experience
#361 opened by maxkazmsft - 0
enable Azure ML pipeline from CSE team - adds AML training pipeline to codebase (models can be trained on AML)
#360 opened by maxkazmsft - 0
The colormap used in mask_to_disk does not support more than 12 classes, will leads to incorrect results for datasets with more than 12 classes
#324 opened by yalaudah - 0
create a backlog of correctness items to fix in the codebase - facilitates correct training pipeline
#329 opened by maxkazmsft - 0
add deconvnet and unet correctness train, val and test metrics as a unit test - facilitates certainty in results
#318 opened by maxkazmsft - 0
update READMEs
#332 opened by fazamani - 0
set clean staging history - facilitates easy release merges of staging into master
#323 opened by maxkazmsft - 0
gradient synthetic dataset does not produce gradient-like images in the train loader - facilitates data correctness
#333 opened by maxkazmsft - 0
Model predictions should not be normalized before saving to disk, leads to incorrect results
#325 opened by yalaudah - 1
Images from the binary dataset result in multiple classes after augmentation
#344 opened by yalaudah - 0
debug test.py not limiting test output in debug mode - facilitates faster build runtimes
#331 opened by maxkazmsft