VisDA-2017

Dataset Information
Modalities
Images
Introduced
2017
License
Custom
Homepage
Official Website
Contents

Overview

VisDA-2017 is a simulation-to-real dataset for domain adaptation with over 280,000 images across 12 categories in the training, validation and testing domains. The training images are generated from the same object under different circumstances, while the validation images are collected from MSCOCO..

Source: Gradually Vanishing Bridge for Adversarial Domain Adaptation
Image Source: http://ai.bu.edu/visda-2017/

Variants: VisDA2017, VisDA-2017

Associated Benchmarks

This dataset is used in 2 benchmarks:

Recent Benchmark Submissions

Task Model Paper Date
Source-Free Domain Adaptation SPM Shuffle PatchMix Augmentation with Confidence-Margin … 2025-05-30
Unsupervised Domain Adaptation TransAdapter TransAdapter: Vision Transformer for Feature-Centric … 2024-12-05
Source-Free Domain Adaptation RCL Empowering Source-Free Domain Adaptation with … 2024-05-28
Source-Free Domain Adaptation SFDA2++ SF(DA)$^2$: Source-free Domain Adaptation Through … 2024-03-16
Source-Free Domain Adaptation SFDA2 SF(DA)$^2$: Source-free Domain Adaptation Through … 2024-03-16
Source-Free Domain Adaptation C-SFDA C-SFDA: A Curriculum Learning Aided … 2023-03-30
Source-Free Domain Adaptation DaC Divide and Contrast: Source-free Domain … 2022-11-12
Source-Free Domain Adaptation NRC Exploiting the Intrinsic Neighborhood Structure … 2021-10-08
Source-Free Domain Adaptation G-SFDA Generalized Source-free Domain Adaptation 2021-08-03
Source-Free Domain Adaptation SHOT++ Source Data-absent Unsupervised Domain Adaptation … 2020-12-14
Source-Free Domain Adaptation SHOT Do We Really Need to … 2020-02-20

Research Papers

Recent papers with results on this dataset:

External Links: