-
Notifications
You must be signed in to change notification settings - Fork 0
Outline of the project
jokteur edited this page Oct 12, 2020
·
1 revision
This page lists the features that have been implemented and features that still need implementation.
- create / import project
- pack data into a single working directory
- name
- description
- working directory
- datasets
- models
- Create dataset
- Import dataset from another project
- Merge datasets
- Split datasets
- Pre-process data
- anonymize dataset : case number, dicom data
- import dicom from folder and specify structure
- import images from folder and specify structure
- apply crop / rotation / resizing on whole dataset
- apply crop from existing model
- apply smart crop : border detection
- filter out
- name
- description
- original anonymized data (non pre-processed)
- pre-processed data
Multiple segmentations can be applied on one dataset
- name
- description
- ML model
- type of segmentation
- resizable circular brush
- lasso select tool
- box select tool
- edge detect brush
- undo / redo
- apply filter on tool (values in HU units, values in RGB)
- apply segmentation from existing model (can be automatic)
- Defines multiple profiles
- Validation levels: modified, partially validated, fully validated
- Image segmentation of image bounding
- Pre-built model
- U-Net model creation : number of layers, input size, output size
- Import external model, pre-trained or not
- Apply on selection of datasets : select level of validation
- Automatic train / test creation
- Data augmentation parameters (shear, crop, resize, rotation)
- Train on raw data or png data (with selection of CT kernel)
- Train with GPU and test memory capacity
- Train with existing weights or retrain from scratch
- Measure area (in m^2, or number of pixels)
- Measure mean value, std deviation
- Histogram of values in mask
- Export csv and select which measurements