Research map / Label-efficient and robust learning / Domain adaptation and generalization
A Compact DNN: Approaching GoogLeNet-Level Accuracy of Classification and Domain Adaptation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use the Conv-M network with dilated convolution and improved deconvolution instead of regular convolution throughout (ablation) | Our DA network · Office-31 A→W | accuracy: 74.2 -> 82.6 | Helped |
| Use the Conv-M network with dilated convolution and improved deconvolution instead of regular convolution throughout (ablation) | ImageNet classification · ImageNet'12 | classification accuracy: 62.2 -> 68.9 | Helped |
| Add G-MMD feature-distribution alignment (ablation) | Our DA (Our net) · Office-31 A→W | accuracy: 76.7 -> 82.6 | Helped |
| Add sample reconstruction (ablation) | Our DA (Our net) · Office-31 D→W | accuracy: 95.4 -> 97.0 | Helped |
| Add sample reconstruction (ablation) | Our DA (Our net) · Office-31 D→A | accuracy: 62.1 -> 67.3 | Helped |
| Add G-MMD feature-distribution alignment (ablation) | Our DA (Our net) · Office-31 W→A | accuracy: 62.0 -> 67.4 | Helped |
| Add sample reconstruction (ablation) | Our DA (Our net) · Office-Caltech A→C | accuracy: 88.7 -> 92.5 | Helped |
| Add G-MMD feature-distribution alignment (ablation) | Our DA (Our net) · Office-Caltech A→C | accuracy: 90.9 -> 92.5 | Helped |
| Use the proposed DA method instead of GRL on the same network | Our network · Office-31 A→W | accuracy: 80.1 -> 82.6 | Helped |
| Use the proposed DA method instead of DAN on the same network | Our network · Office-Caltech A→C | accuracy: 91.6 -> 92.5 | Helped |
| Use the proposed compact network instead of GoogLeNet with the proposed DA method | Our DA · Office-31 A→W | accuracy: 83.0 -> 82.6 | Hurt |
Closest papers by meaning
- Boosting Domain Adaptation by Discovering Latent Domains (CVPR 2018 · similarity 0.87)
- GCAN: Graph Convolutional Adversarial Network for Unsupervised Domain Adaptation (CVPR 2019 · similarity 0.87)
- Feature-Critic Networks for Heterogeneous Domain Generalization (ICML 2019 · similarity 0.86)
- Image to Image Translation for Domain Adaptation (CVPR 2018 · similarity 0.86)
- Adversarial Discriminative Domain Adaptation (CVPR 2017 · similarity 0.86)
- Improving Domain-Specific Classification by Collaborative Learning with Adaptation Networks (AAAI 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.