Research map / Continual learning / Class-incremental
Knowledge Transfer with Interactive Learning of Semantic Relationships
Read the paperCheck 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 semantic constraints from interactive queries (Interactive vs LME-Transfer) | LME-Transfer · AWA 2/class | classification accuracy (%): 24.59±2.23 -> 27.24±1.82 | Helped |
| Add knowledge transfer from anchor classes (LME-Transfer vs LME) | LME · AWA 2/class | classification accuracy (%): 22.51±2.48 -> 24.59±2.23 | Helped |
| Interactive vs batch-mode Active scoring (ablation) | Active · AWA 2/class | classification accuracy (%): 26.62±1.67 -> 27.24±1.82 | Helped |
| Interactive vs batch-mode Active scoring (ablation) | Active · AWA 5/class | classification accuracy (%): 32.42±1.45 -> 33.31±1.28 | Helped |
| Interactive vs batch-mode Active scoring (ablation) | Active · ImageNet-50 5/class | classification accuracy (%): 28.55±2.07 -> 29.08±1.88 | Mixed |
| Interactive vs batch-mode Active scoring (ablation) | Active · ImageNet-50 10/class | classification accuracy (%): 35.60±1.01 -> 35.62±1.01 | No clear effect |
| Interactive vs Random query ordering (ablation) | Random · AWA 2/class | classification accuracy (%): 24.75±2.11 -> 27.24±1.82 | Helped |
| Interactive vs Entropy-based scoring (ablation) | Entropy · AWA 2/class | classification accuracy (%): 24.96±2.24 -> 27.24±1.82 | Helped |
| Interactive vs Entropy-based scoring (ablation) | Entropy · ImageNet-50 10/class | classification accuracy (%): 35.64±0.99 -> 35.62±1.01 | No clear effect |
| Active-Regression instead of Active (regress accuracy) (ablation) | Active · ImageNet-50 2/class | classification accuracy (%): 24.35±2.42 -> 23.34±2.76 | Hurt |
| Interactive vs oracle upper bound scored on test set (ablation) | Interactive · AWA 2/class | classification accuracy (%): 27.24±1.82 -> 28.57±1.85 | Helped |
| Interactive vs oracle upper bound scored on test set (ablation) | Interactive · AWA 10/class | classification accuracy (%): 36.46±1.60 -> 36.86±1.83 | No clear effect |
Closest papers by meaning
- Revisiting Knowledge Transfer for Training Object Class Detectors (CVPR 2018 · similarity 0.84)
- Category Query Learning for Human-Object Interaction Classification (CVPR 2023 · similarity 0.83)
- Semantic Relation Reasoning for Shot-Stable Few-Shot Object Detection (CVPR 2021 · similarity 0.83)
- Semantic-Guided Novel Category Discovery (AAAI 2024 · similarity 0.82)
- Cyclic Contrastive Knowledge Transfer for Open-Vocabulary Object Detection (ICLR 2025 · similarity 0.82)
- Enhancing Visual Continual Learning with Language-Guided Supervision (CVPR 2024 · similarity 0.81)
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.