Purpose
What SegCraft does.
Aim
Use one setup to train a segmentation model, compare it, and run it on images or videos.
What it does
Layered YAML configuration drives consistent model, data, evaluation, and inference contracts across CLI, Python, and web interfaces with optional model backends.
Good at
- Model presets
- Comparative evaluation
- Video workflows
Flow
The SegCraft flow.
- 01
Configure
Merge a base configuration, a reusable preset, and a local override.
- 02
Validate
Check the environment, task shape, model backend, and device visibility.
- 03
Train
Run a consistent model, data, schedule, and metrics configuration.
- 04
Evaluate
Measure the same configured task without rebuilding the pipeline.
- 05
Infer
Produce masks, overlays, comparison video, and a machine-readable summary.
Try it
