Finetune CT segmentation with checkpoint evidence.
What it does
Finetunes NV-Segment-CT VISTA3D on labeled CT scans and returns checkpoints with evaluation evidence.
When to use it
Use for smoke checks or dataset finetuning, choosing softmax for fixed, mutually exclusive classes and the standard workflow for point prompts or dynamically selected classes; not for clinical validation.
How to use it
Give it a labeled CT dataset and a requested training run; it produces a finetuned checkpoint and result JSON in the output directory.
What you provide
- Your files
Uses
NS
Good to know
- cannot be undone: bundle/configs/
Latest review
What you need · 13
Requires Python 3.10 or later.
Requires Torch, with CUDA support for GPU runs.
Access · 8
This skill
bundle/configs/
Write
Overwrites bundled training configs (irreversible)
MLflow
Write
Sends training tracking data