Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Introducing Intermediate Domains for Effective Self-Training during Test-Time

About

Experiencing domain shifts during test-time is nearly inevitable in practice and likely results in a severe performance degradation. To overcome this issue, test-time adaptation continues to update the initial source model during deployment. A promising direction are methods based on self-training which have been shown to be well suited for gradual domain adaptation, since reliable pseudo-labels can be provided. In this work, we address two problems that exist when applying self-training in the setting of test-time adaptation. First, adapting a model to long test sequences that contain multiple domains can lead to error accumulation. Second, naturally, not all shifts are gradual in practice. To tackle these challenges, we introduce GTTA. By creating artificial intermediate domains that divide the current domain shift into a more gradual one, effective self-training through high quality pseudo-labels can be performed. To create the intermediate domains, we propose two independent variations: mixup and light-weight style transfer. We demonstrate the effectiveness of our approach on the continual and gradual corruption benchmarks, as well as ImageNet-R. To further investigate gradual shifts in the context of urban scene segmentation, we publish a new benchmark: CarlaTTA. It enables the exploration of several non-stationary domain shifts.

Robert A. Marsden, Mario D\"obler, Bin Yang• 2022

Related benchmarks

TaskDatasetResultRank
Image ClassificationImageNet-R (test)--
105
Image ClassificationCIFAR-10C Severity Level 5 (test)
Average Error Rate (Severity 5)15.6
62
Image ClassificationImageNet-C level 5
Avg Top-1 Acc (ImageNet-C L5)57.1
61
Image ClassificationCIFAR-100C Level 5 (test)--
45
Image ClassificationCIFAR10C level 1-5
Error Rate11.8
18
Image ClassificationImageNet-C level 1-5
Error Rate (L1-5)37.7
9
Image ClassificationCIFAR100C level 5
Classification Error Rate26.1
9
Image ClassificationCIFAR10C level 5
Error Rate0.13
9
Image ClassificationDomainNet 126 (test)
Accuracy (real)38.7
9
Showing 9 of 9 rows

Other info

Follow for update