Key Features and Enhancements
The DIGITS version 6.0 release includes the following key features and enhancements. These enhancements are new since version DIGITS version 5.0.
- Added TensorFlow™ support for DIGITS
- Added TensorBoard support for DIGITS
- Updated documentation for how to use TensorFlow inside DIGITS
- Added Siamese Network examples in DIGITS with TensorFlow, NVCaffe™ , and Torch™
- Updated auto-encoder, binary-segmentation, fine-tuning, and regression with TensorFlow support
- Added LeNet, AlexNet, and GoogLeNet template models for TensorFlow
- Added GANS Model in the Model store
- Added U-Net Model in Model store
- Added tutorial for DICOM data ingestion
- Added support for URL prefix
- Fixed issues with loading and saving TensorFlow models
Using DIGITS 6.0
Ensure you are familiar with TensorFlow, TensorFlow API, and the TensorFlow workflow. For more information, see the TensorFlow User Guide.
Known Issues
- TensorFlow models were found to be taking slightly longer to train than Torch and NVCaffe.
- GANS cannot take advantage of multi-GPU set-ups.