Research map / Label-efficient and robust learning / Domain adaptation and generalization
AutoDIAL: Automatic DomaIn Alignment Layers
Fabio Maria Carlucci, Lorenzo Porzi, Barbara Caputo, Elisa Ricci, Samuel Rota Bulo
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add learned alpha (instead of alpha fixed to 1) in DA-layers (ablation) | AutoDIAL (AlexNet) · Office31 | accuracy | Improved |
| add entropy loss alone to source-only training | AlexNet · Office31 (D->A, W->A) | accuracy | Hurt |
| add DA-layers together with the entropy loss (ablation) | AlexNet · Office31 | accuracy | Improved |
| replace AdaBN-style independent alignment with learned alpha plus entropy regularization (ablation) | Inception-BN · Office31 (average) | accuracy (%): 76.7 -> 82.2 | Improved |
| replace AdaBN with AutoDIAL | Inception-BN · Cross-Dataset Testbed Caltech->ImageNet (setting 1) | accuracy (%): 82.2 -> 85.2 | Improved |
| replace AdaBN with AutoDIAL | Inception-BN · Cross-Dataset Testbed ImageNet->Caltech (setting 1) | accuracy (%): 87.3 -> 90.5 | Improved |
| replace AdaBN with AutoDIAL | Inception-BN · Cross-Dataset Testbed Caltech->ImageNet (setting 2) | accuracy (%): 81.9 -> 84.2 | Improved |
| replace AdaBN with AutoDIAL | Inception-BN · Cross-Dataset Testbed ImageNet->Caltech (setting 2) | accuracy (%): 86.5 -> 89.8 | Improved |
| apply AdaBN without target-aware training | Inception-BN · Cross-Dataset Testbed ImageNet->Caltech (setting 1) | accuracy (%): 88.4 -> 87.3 | Hurt |
| add domain alignment (AutoDIAL) to source-only network | AlexNet · Office31 (average) | accuracy (%): 70.1 -> 77.1 | Improved |
| add domain alignment (AutoDIAL) to source-only network | AlexNet · Office-Caltech (average) | accuracy (%): 84.0 -> 90.3 | Improved |
| AutoDIAL versus RTN | AlexNet · Office-Caltech (average) | accuracy (%): 90.6 vs 90.3 | No clear effect |
Later papers that compared with this method
2 later accepted papers report AutoDIAL as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Domain2Vec: Domain Embedding for Unsupervised Domain Adaptation (ECCV 2020 · better on 2 of 2 shared results)
- Learning Semantic Representations for Unsupervised Domain Adaptation (ICML 2018 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MultiDIAL: Domain Alignment Layers for (Multisource) Unsupervised Domain Adaptation (TPAMI 2020 · similarity 0.90)
- Co-regularized Alignment for Unsupervised Domain Adaptation (NeurIPS 2018 · similarity 0.89)
- Unsupervised Domain Adaptation via Regularized Conditional Alignment (ICCV 2019 · similarity 0.88)
- ToAlign: Task-Oriented Alignment for Unsupervised Domain Adaptation (NeurIPS 2021 · similarity 0.88)
- Exploiting Local Feature Patterns for Unsupervised Domain Adaptation (AAAI 2019 · similarity 0.88)
- Boosting Domain Adaptation by Discovering Latent Domains (CVPR 2018 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
