# Weakly supervised training of universal visual concepts for multi-domain semantic segmentation

Source: [Weakly supervised training of universal visual concepts for multi-domain semantic segmentation](<https://www.sota2.com/research/paper/68fa72751e209131cea56815>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-28T13:57:04.875Z

Authors: Petra Bevandi\\'c, Marin Or\\v{s}i\\'c, Ivan Grubi\\v{s}i\\'c, Josip \\v{S}ari\\'c, Sini\\v{s}a \\v{S}egvi\\'c

Published: 2022-12-20T15:25:38.000Z

## Abstract

Deep supervised models have an unprecedented capacity to absorb large quantities of training data. Hence, training on multiple datasets becomes a method of choice towards strong generalization in usual scenes and graceful performance degradation in edge cases. Unfortunately, different datasets often have incompatible labels. For instance, the Cityscapes road class subsumes all driving surfaces, while Vistas defines separate classes for road markings, manholes etc. Furthermore, many datasets have overlapping labels. For instance, pickups are labeled as trucks in VIPER, cars in Vistas, and vans in ADE20k. We address this challenge by considering labels as unions of universal visual concepts. This allows seamless and principled learning on multi-domain dataset collections without requiring any relabeling effort. Our method achieves competitive within-dataset and cross-dataset generalization, as well as ability to learn visual concepts which are not separately labeled in any of the training datasets. Experiments reveal competitive or state-of-the-art performance on two multi-domain dataset collections and on the WildDash 2 benchmark.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Semantic Segmentation on SUN](<https://www.sota2.com/research/sota/semantic-segmentation-on-sun.md>): Semantic segmentation; SUN; mIoU: 46.9.

- [Semantic Segmentation on WildDash bench (test)](<https://www.sota2.com/research/sota/semantic-segmentation-on-wilddash-bench-test.md>): Semantic segmentation; WildDash bench (test); mIoU Meta Avg (cla): 46.9.

- [Semantic Segmentation on IDD](<https://www.sota2.com/research/sota/semantic-segmentation-on-idd.md>): Semantic segmentation; IDD; mIoU: 56.7.

- [Semantic Segmentation on BDD100K](<https://www.sota2.com/research/sota/semantic-segmentation-on-bdd100k.md>): Semantic segmentation; BDD100K; mIoU: 60.4.

- [Semantic Segmentation on Mapillary](<https://www.sota2.com/research/sota/semantic-segmentation-on-mapillary.md>): Semantic segmentation; Mapillary; mIoU: 44.2.

- [Semantic Segmentation on COCO (mIoU)](<https://www.sota2.com/research/sota/semantic-segmentation-on-coco-miou.md>): Semantic segmentation; COCO; mIoU: 39.3.

- [Semantic Segmentation on Cityscapes](<https://www.sota2.com/research/sota/semantic-segmentation-on-cityscapes.md>): Semantic segmentation; Cityscapes; mIoU: 76.1.

- [Semantic Segmentation on ADE20K](<https://www.sota2.com/research/sota/semantic-segmentation-on-ade20k.md>): Semantic segmentation; ADE20K; mIoU: 35.6.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2212.10340>)

[PDF](<https://arxiv.org/pdf/2212.10340.pdf>)