Executable Notebooks

These notebooks are maintained as user-facing tutorials. They are intended to be run from top to bottom and to show the exact point where RankSEG replaces a standard segmentation prediction step.

Notebook	What it teaches	Best entry point
quickstart.ipynb	A short PyTorch walkthrough with a pretrained DeepLabV3-style workflow: model output, probability map, baseline mask, RankSEG mask.	New users who want the fastest runnable example.
rankseg_with_transformers.ipynb	A Hugging Face tutorial that first runs the official baseline and then changes only the final post-processing step to RankSEG.	Users with processor -> model -> outputs code.
rankseg_with_sam_family.ipynb	A SAM-family tutorial covering SAM1, SAM2, and SAM3 adapters, including geometry restoration before RankSEG is called.	Users working with prompt, instance, or semantic masks from SAM-family models.
rankseg_with_paddleseg.ipynb	A PaddleSeg-oriented walkthrough that shows the probability conversion step before calling the PyTorch-based RankSEG predictor.	PaddleSeg users evaluating the external/community-maintained path.

Colab links

• Open quickstart.ipynb in Colab

• Open rankseg_with_transformers.ipynb in Colab

• Open rankseg_with_sam_family.ipynb in Colab

• Open rankseg_with_paddleseg.ipynb in Colab

How to read the notebooks with the docs

Use Official Integrations first to choose the correct backend. Then open the matching notebook when you want a runnable, visual workflow. The conceptual pages define the tensor shapes, probability conventions, option names, and supported output families; the notebooks show those contracts in complete inference code.