seamseg

Research code for panoptic segmentation, which labels every pixel of an image by class and by individual object. It is a neural network architecture from a CVPR paper.

Browse open issues

Pick a small issue you can reproduce, then fix it with your AI coding agent.

Share on XLicense: BSD-3-Clause

Overview

Seamless Scene Segmentation is a CNN-based architecture for panoptic segmentation, which gives every pixel of an image both a class label and an instance label. It combines multi-scale features from a Feature Pyramid Network with context from a light DeepLab-like module. The repository holds PyTorch training and evaluation code, built on a re-implementation of Mask R-CNN, and comes from a CVPR 2019 paper.

Key features

  • Panoptic segmentation with class and instance labels per pixel
  • Segmentation head joining Feature Pyramid Network features with DeepLab-like context
  • PyTorch training and evaluation code
  • Based on a re-implementation of Mask R-CNN

Best for

Researchers who want to train or evaluate panoptic segmentation. The stated setup is CUDA 10.1, Linux with GCC 7 or 8, and PyTorch 1.1.0.

Upstream
mapillary/seamseg
Fork on GitHub
Guo-astro/seamseg
Upstream stars
301
Category
Finance, data and research
Language
Python
License
BSD-3-Clause
Forked
2020-12-05
Sync status
In syncLast synced 2026-10-19

More in Finance, data and research