Skip to content

Commit

Permalink
v1.9
Browse files Browse the repository at this point in the history
  • Loading branch information
Romain-Laine committed Aug 7, 2020
1 parent 8db8f0e commit 057412d
Show file tree
Hide file tree
Showing 15 changed files with 14 additions and 12 deletions.
Binary file modified .DS_Store
Binary file not shown.
Binary file modified Colab_notebooks/.DS_Store
Binary file not shown.
2 changes: 1 addition & 1 deletion Colab_notebooks/CARE_2D_ZeroCostDL4Mic.ipynb

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion Colab_notebooks/CARE_3D_ZeroCostDL4Mic.ipynb

Large diffs are not rendered by default.

2 changes: 2 additions & 0 deletions Colab_notebooks/ChangeLog.txt
Original file line number Diff line number Diff line change
Expand Up @@ -19,12 +19,14 @@ Main notebooks:
——————————————

- StarDist 2D, StarDist 3D, CARE 2D and CARE 3D notebooks now uses TensorFlow 2.2 (instead of TF 1.5.15)

- YOLOv2 notebook: QC section now uses the same mAP function as the training, for better consistency of results; ground-truth labels and predicted labels in the QC section are exported to the QC folder as csv files which holde the bounding box coordinates and class labels; Display of prediction results now more consistent with display of GT labels; Updated Augmentation Section, now between 2-8 times augmentation of dataset possible; Additional csv file with predicted bounding box coordinates in a format suitable for use in imageJ as Results Table is now exported to the user's results folder in the Prediction section; Added 'training_times' as hyperparameter for improved tuning of model training; Tracking of mAP during training implemented; After training, model with best validation performance, best mAP score and the last model weights are saved to allow easier performance comparison by the user; Updated explanation of parameters and QC section.

- 3D U-Net:
1. Added ability to train network on non-binary targets
2. Added ability to choose loss, metrics and optimizer
3. Fixed data generator bug leading to erroneous generator length when choosing random_crop
4. Added support for using the imgaug library and creating custom augmentation pipelines

+ minor modifications and bug fixes

Expand Down
2 changes: 1 addition & 1 deletion Colab_notebooks/CycleGAN_ZeroCostDL4Mic.ipynb

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion Colab_notebooks/Deep-STORM_2D_ZeroCostDL4Mic.ipynb

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion Colab_notebooks/Noise2VOID_2D_ZeroCostDL4Mic.ipynb

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion Colab_notebooks/Noise2VOID_3D_ZeroCostDL4Mic.ipynb

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion Colab_notebooks/Stardist_2D_ZeroCostDL4Mic.ipynb

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion Colab_notebooks/Stardist_3D_ZeroCostDL4Mic.ipynb

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion Colab_notebooks/U-net_2D_ZeroCostDL4Mic.ipynb

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion Colab_notebooks/YOLOv2_ZeroCostDL4Mic.ipynb

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion Colab_notebooks/fnet_ZeroCostDL4Mic.ipynb

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion Colab_notebooks/pix2pix_ZeroCostDL4Mic.ipynb

Large diffs are not rendered by default.

0 comments on commit 057412d

Please sign in to comment.