Research map / Label-efficient and robust learning / Domain adaptation and generalization
Visual Recognition in RGB Images and Videos by Learning from RGB-D Data
Wen Li, Lin Chen, Dong Xu, Luc Van Gool
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 depth features in the source domain plus MMD domain adaptation (DAM2S_A) | SVM_A · RGB-D Object to Caltech-256 object recognition | accuracy (%): 27.52 -> 53.17 | Improved |
| add privileged depth information via SVM2K without domain adaptation | SVM_A · RGB-D Object to Caltech-256 object recognition | accuracy (%): 27.52 -> 34.21 | Improved |
| add depth features and domain adaptation instead of best DA baseline DIP | DIP · RGB-D Object to Caltech-256 object recognition | accuracy (%): 45.13 -> 56.89 | Improved |
| use domain adaptation DAM2S_C on cross-dataset action recognition | SVM_A · Hollywood 3D to Hollywood2 cross-dataset action recognition | accuracy (%): 19.17 -> 39.17 | Improved |
| use multi-source domain adaptation DAM instead of SVM_A | SVM_A · Hollywood 3D to Hollywood2 cross-dataset action recognition | accuracy (%): 19.17 -> 19.32 | No clear effect |
| use DAM2S_A on cross-view action recognition | SVM_A · ACT42 cross-view action recognition | accuracy (%): 62.33 -> 68.86 | Improved |
| use SA-based regularizer (DAM2S_C) instead of MMD regularizer (DAM2S_A) under cross-view (ablation) | DAM2S · ACT42 cross-view action recognition | accuracy (%): 68.86 -> 65.56 | Hurt |
| use SA-based regularizer (DAM2S_C) instead of MMD regularizer (DAM2S_A) cross-dataset (ablation) | DAM2S · Hollywood 3D to Hollywood2 cross-dataset action recognition | accuracy (%): 30.64 -> 39.17 | Improved |
| use SVM2K subproblem (DAM2S_B) instead of SVM subproblem (DAM2S_A) (ablation) | DAM2S · Hollywood 3D to Hollywood2 cross-dataset action recognition | training time (s): 20.7 -> 168.4 | Hurt |
| replace DAM2S_A with DAM2S_C (ablation) | DAM2S · Hollywood 3D to Hollywood2 cross-dataset action recognition | training time (s): 20.7 -> 398.2 | Hurt |
| use SVM2K (exploit depth with two classifiers) instead of KCCA | multi-view baselines · RGB-D Object to Caltech-256 object recognition | accuracy (%): 31.52 -> 34.21 | Improved |
Closest papers by meaning
- Joint Geometrical and Statistical Alignment for Visual Domain Adaptation (CVPR 2017 · similarity 0.88)
- Multivariate Regression on the Grassmannian for Predicting Novel Domains (CVPR 2016 · similarity 0.87)
- Domain Generalization and Adaptation Using Low Rank Exemplar SVMs (TPAMI 2017 · similarity 0.87)
- LEARNING FACTORIZED REPRESENTATIONS FOR OPEN-SET DOMAIN ADAPTATION (ICLR 2019 · similarity 0.86)
- Domain2Vec: Domain Embedding for Unsupervised Domain Adaptation (ECCV 2020 · similarity 0.86)
- Learning with Privileged Information via Adversarial Discriminative Modality Distillation (TPAMI 2019 · similarity 0.86)
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.
