Skip to main content

PyTorch Segmentation Models Trainer

Framework for training semantic segmentation models with PyTorch Lightning and Hydra

Easy to Use

Configuration-Driven

Define your entire training pipeline with YAML files. Use Hydra for composable, reproducible experiments. Build configs visually with the Config Builder.

Focus on What Matters

Multiple Model Types

Support for semantic segmentation, frame field models, object detection, instance segmentation, and Polygon-RNN — all through a unified interface.

Powered by React

Advanced Polygonization

Convert segmentation masks into precise vector polygons using active skeletons, active contours, simple polygonization, or Polygon-RNN.