On the Road to Online Adaptation for Semantic Image Segmentation

Riccardo Volpi, Diane Larlus, Gabriela Csurka

Code that ran

We have not lifted any functions out of this paper's repositories yet, so there is nothing we have run. If it links a repository, it is listed below.

Repositories linked to this paper

naver/oasis — marked official in Papers with Code; named in the paper

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

Abstract

Figure 1. The proposed OASIS benchmark. We formalize a domain adaptation task which requires online, unsupervised adaptation of semantic segmentation models and propose a novel benchmark to tackle it. It is composed of three steps. Train: A model is trained offline on simulated data (top-left); Val: Several adaptation strategies are validated on simulated data organized in sequentially shifting domains (e.g. , sunny-to-rainy, highway-to-city), to mimic deploy (top-right). Deploy: The best validated strategy is applied to the test set (bottom).

For agents

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

get_harvested_code_for_paper("2203.16195")
get_code_for_paper("2203.16195")
have("2203.16195")

Connect an agent — have() is free.