File Download
There are no files associated with this item.
Supplementary
-
Citations:
- Appears in Collections:
Conference Paper: Back to the Source: Diffusion-Driven Test-Time Adaptation
Title | Back to the Source: Diffusion-Driven Test-Time Adaptation |
---|---|
Authors | |
Issue Date | 18-Jun-2023 |
Abstract | Test-time adaptation harnesses test inputs to improve the accuracy of a model trained on source data when tested on shifted target data. Existing methods update the source model by (re-)training on each target domain. While effective, re-training is sensitive to the amount and order of the data and the hyperparameters for optimization. We instead update the target data, by projecting all test inputs toward the source domain with a generative diffusion model. Our diffusion-driven adaptation method, DDA, shares its models for classification and generation across all domains. Both models are trained on the source domain, then fixed during testing. We augment diffusion with image guidance and self-ensembling to automatically decide how much to adapt. Input adaptation by DDA is more robust than prior model adaptation approaches across a variety of corruptions, architectures, and data regimes on the ImageNet-C benchmark. With its input-wise updates, DDA succeeds where model adaptation degrades on too little data in small batches, dependent data in non-uniform order, or mixed data with multiple corruptions. |
Persistent Identifier | http://hdl.handle.net/10722/333874 |
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Gao, Jin | - |
dc.contributor.author | Zhang, Jialing | - |
dc.contributor.author | Liu, Xihui | - |
dc.contributor.author | Darrell, Trevor | - |
dc.contributor.author | Shelhamer, Evan | - |
dc.contributor.author | Wang, Dequan | - |
dc.date.accessioned | 2023-10-06T08:39:48Z | - |
dc.date.available | 2023-10-06T08:39:48Z | - |
dc.date.issued | 2023-06-18 | - |
dc.identifier.uri | http://hdl.handle.net/10722/333874 | - |
dc.description.abstract | <p>Test-time adaptation harnesses test inputs to improve the accuracy of a model trained on source data when tested on shifted target data. Existing methods update the source model by (re-)training on each target domain. While effective, re-training is sensitive to the amount and order of the data and the hyperparameters for optimization. We instead update the target data, by projecting all test inputs toward the source domain with a generative diffusion model. Our diffusion-driven adaptation method, DDA, shares its models for classification and generation across all domains. Both models are trained on the source domain, then fixed during testing. We augment diffusion with image guidance and self-ensembling to automatically decide how much to adapt. Input adaptation by DDA is more robust than prior model adaptation approaches across a variety of corruptions, architectures, and data regimes on the ImageNet-C benchmark. With its input-wise updates, DDA succeeds where model adaptation degrades on too little data in small batches, dependent data in non-uniform order, or mixed data with multiple corruptions.<br></p> | - |
dc.language | eng | - |
dc.relation.ispartof | 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (17/06/2023-24/06/2023, Vancouver, BC, Canada) | - |
dc.title | Back to the Source: Diffusion-Driven Test-Time Adaptation | - |
dc.type | Conference_Paper | - |
dc.identifier.doi | 10.48550/arXiv.2207.03442 | - |