Skip to content

Revisiting K-Net for Real-Time Panoptic Segmentation. Code release for our IV 2023 paper.

License

Notifications You must be signed in to change notification settings

uulm-mrm/RT-K-Net

Repository files navigation

RT-K-Net: Revisiting K-Net for Real-Time Panoptic Segmentation

This repository contains the official implementation of our IV 2023 paper RT-K-Net: Revisiting K-Net for Real-Time Panoptic Segmentation.

Installation

See INSTALL.md for instructions on how to prepare your environment to use RT-K-Net.

Usage

See datasets/README.md for instructions on how to prepare datasets for RT-K-Net.

See GETTING_STARTED.md for instructions on how to train and evaluate models, or run inference on demo images.

Model Zoo

The model files provided below are made available under the CC BY-NC-SA 4.0 license. The Cityscapes model was trained using 4 NVIDIA 2080Ti GPUs, the Mapillary Vistas model was trained using 8 NVIDIA 2080Ti GPUs.

Name PQ PQ_St PQ_Th Download
RT-K-Net Cityscapes Fine 60.2 66.5 51.5 model
RT-K-Net Mapillary Vistas 33.2 45.8 23.6 model

Reference

Please use the following citations when referencing our work:

RT-K-Net: Revisiting K-Net for Real-Time Panoptic Segmentation (IV 2023)
Markus Schön, Michael Buchholz and Klaus Dietmayer, [arxiv]

@InProceedings{Schoen_2023_IV,
    author    = {Sch{\"o}n, Markus and Buchholz, Michael and Dietmayer, Klaus},
    title     = {RT-K-Net: Revisiting K-Net for Real-Time Panoptic Segmentation},
    booktitle = {IEEE Intelligent Vehicles Symposium},
    year      = {2023}
}

Acknowledgement

We used and modified code parts from other open source projects, we especially like to thank the authors of:

About

Revisiting K-Net for Real-Time Panoptic Segmentation. Code release for our IV 2023 paper.

Topics

Resources

License

Stars

Watchers

Forks