Implementation of 2D and 3D cnn classification workflow
Implemented a classification training and classification inference workflows.
classification training
- Trains Resnet18 (2D or 3D)
- Adjustable parameters via UI: used tags, task (binary, multiclass, multilabel), dimensions (2D or 3D), patch size, batch size and epochs
- Z-Score normalization
- batchgenerators
- Tensorboard
classification inference
- Takes model from training and predicts tag
- Writes tag in OpenSearch