Research map / Detection and segmentation / CNN features
Feature Space Transfer for Data Augmentation
Bo Liu, Xudong Wang, Mandar Dixit, Roland Kwitt, Nuno Vasconcelos
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 |
|---|---|---|---|
| Augment one-shot training features with pose- and depth-conditioned FATTEN features instead of using only the single-instance baseline | Linear SVM · SUN-RGBD T1 one-shot | accuracy: 33.74 -> 44.99 | Improved |
| Augment one-shot training features with pose- and depth-conditioned FATTEN features instead of using only the single-instance baseline | Linear SVM · SUN-RGBD T2 one-shot | accuracy: 23.76 -> 34.70 | Improved |
| Augment one-shot training features with pose- and depth-conditioned FATTEN features instead of using only the single-instance baseline | Linear SVM · SUN-RGBD T3 one-shot | accuracy: 22.84 -> 32.20 | Improved |
| Use FATTEN feature augmentation rather than AGA augmentation | One-shot recognition · SUN-RGBD T1 | accuracy: 39.10 -> 44.99 | Improved |
| Use FATTEN feature augmentation rather than AGA augmentation | One-shot recognition · SUN-RGBD T2 | accuracy: 30.12 -> 34.70 | Improved |
| Use FATTEN feature augmentation rather than AGA augmentation | One-shot recognition · SUN-RGBD T3 | accuracy: 26.67 -> 32.20 | Improved |
| Augment five-shot training features with FATTEN pose- and depth-conditioned features | Linear SVM · SUN-RGBD T1 five-shot | accuracy: 50.03 -> 58.82 | Improved |
| Augment five-shot training features with FATTEN pose- and depth-conditioned features | Linear SVM · SUN-RGBD T2 five-shot | accuracy: 36.76 -> 50.69 | Improved |
| Augment five-shot training features with FATTEN pose- and depth-conditioned features | Linear SVM · SUN-RGBD T3 five-shot | accuracy: 37.37 -> 47.07 | Improved |
| Generate features for varied poses rather than retrieve using real features | ModelNet retrieval · ModelNet test | pose retrieval mAP: 54.58 -> 77.62 | Improved |
| Generate features for varied poses rather than retrieve using real features | ModelNet retrieval · ModelNet test | category retrieval mAP: 32.71 -> 28.89 | Hurt |
| Generate features for varied poses rather than retrieve using real features | ModelNet retrieval · ModelNet test | joint pose and category retrieval mAP: 23.65 -> 11.07 | Hurt |
Closest papers by meaning
- Pose Augmentation: Class-agnostic Object Pose Transformation for Object Recognition (ECCV 2020 · similarity 0.86)
- Interpretable Transformations With Encoder-Decoder Networks (ICCV 2017 · similarity 0.85)
- Transferring Vision-Language Models for Visual Recognition: A Classifier Perspective (IJCV 2023 · similarity 0.85)
- Big Transfer (BiT): General Visual Representation Learning (ECCV 2020 · similarity 0.85)
- Features are fate: a theory of transfer learning in high-dimensional regression (ICML 2025 · similarity 0.84)
- CAP-Net: A Unified Network for 6D Pose and Size Estimation of Categorical Articulated Parts from a Single RGB-D Image (CVPR 2025 · similarity 0.83)
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.
