Research map / Detection and segmentation / Open-set and few-shot detection
AGA: Attribute-Guided Augmentation
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 |
|---|---|---|---|
| Add depth-guided synthesized features | One-shot object recognition with Fast R-CNN FC7 features · SUN RGB-D target sets T1, T2, and T3 | accuracy: 33.74 -> 38.32; 23.76 -> 28.49; 22.84 -> 25.52 | Helped |
| Add pose-guided synthesized features | One-shot object recognition with Fast R-CNN FC7 features · SUN RGB-D target sets T1, T2, and T3 | accuracy: 33.74 -> 37.25; 23.76 -> 27.15; 22.84 -> 24.34 | Helped |
| Add both depth- and pose-guided synthesized features | One-shot object recognition with Fast R-CNN FC7 features · SUN RGB-D target sets T1, T2, and T3 | accuracy: 33.74 -> 39.10; 23.76 -> 30.12; 22.84 -> 26.67 | Helped |
| Add both depth- and pose-guided synthesized features | Five-shot object recognition with Fast R-CNN FC7 features · SUN RGB-D target sets T1, T2, and T3 | accuracy: 50.03 -> 56.92; 36.76 -> 47.04; 37.37 -> 42.87 | Helped |
| Augment detected object features with depth-guided features and form Fisher vectors | RCNN max-pooling scene classifier · MIT Indoor one-shot, 25 classes | accuracy: 13.97 -> 15.13 | Helped |
| Augment detected object features with pose-guided features and form Fisher vectors | RCNN max-pooling scene classifier · MIT Indoor one-shot, 25 classes | accuracy: 13.97 -> 14.63 | Helped |
| Combine AGA Fisher vectors with max pooling | RCNN scene classifier · MIT Indoor one-shot, 25 classes | accuracy: 13.97 -> 16.04; 13.97 -> 15.52; 13.97 -> 16.32 | Helped |
| Add AGA-augmented features | Semantic FV scene classifier · MIT Indoor one-shot, 25 classes | accuracy: 32.75 -> 34.36 | Helped |
| Combine AGA features with Places CNN features | Places scene classifier · MIT Indoor one-shot, 25 classes | accuracy: 51.28 -> 52.11 | Helped |
| Train attribute regression object-agnostically rather than separately per object | AGA attribute regressor · SUN RGB-D, 19 objects | depth MAE [m]: 0.39 -> 0.51 | Hurt |
| Train attribute regression object-agnostically rather than separately per object | AGA attribute regressor · SUN RGB-D, 19 objects | pose MAE [deg]: 40.33 -> 41.12 | Hurt |
| Use a two-layer attribute regressor rather than adding more layers (ablation) | AGA attribute regressor · SUN RGB-D attribute prediction | attribute prediction performance | No clear effect |
Closest papers by meaning
- A Data-Augmentation Is Worth A Thousand Samples: Analytical Moments And Sampling-Free Training (NeurIPS 2022 · similarity 0.86)
- TeachAugment: Data Augmentation Optimization Using Teacher Knowledge (CVPR 2022 · similarity 0.85)
- AutoAugment: Learning Augmentation Strategies From Data (CVPR 2019 · similarity 0.85)
- Soft Augmentation for Image Classification (CVPR 2023 · similarity 0.85)
- Black Sheep in the Herd: Playing with Spuriously Correlated Attributes for Vision-Language Recognition (ICLR 2025 · similarity 0.84)
- Super-Class Guided Transformer for Zero-Shot Attribute Classification (AAAI 2025 · similarity 0.84)
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.