SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2108.06962 · ICCV · 2021

Multi-Target Adversarial Frameworks for Domain Adaptation in Semantic Segmentation

Patrick Pérez, Matthieu Cord, Tuan-Hung Vu, Antoine Saporta

arXiv · PDF · Open in the Atlas

Code that ran

We have not lifted any functions out of this paper's repositories yet, so there is nothing we have run. The repositories linked to it are listed below.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Multi-target unsupervised domain adaptation (UDA) for semantic segmentation. In the standard single-target setting, UDA methods produce good segmentation in the target domain they are trained on, but generalize poorly to other unseen domains. Multi-target UDA aims at excelling in the multiple domains the model is trained on. (Top) The information available during training is composed of source-domain RGB images with ground-truth semantic maps (green), here from GTA5, and unannotated RGB images from target domain(s) (blue), here from IDD ('target 1') and Cityscapes ('target T '). (Bottom) Test-time segmentation is on new images from the target domains, without knowing which domain they stem from.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2108.06962")
get_code_for_paper("2108.06962")
have("2108.06962")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2108.06962.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2108.06962.svg)](https://syntology.ai/paper/2108.06962)

Connect an agent — have() is free.