Research map / Vision-language and multimodal / Cross-modal learning
Cross-Task Transfer for Geotagged Audiovisual Aerial Scene Recognition
Di Hu, Xuhong Li, Lichao Mou, Pu Jin, Dong Chen, Liping Jing, Xiaoxiang Zhu, Dejing Dou
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 |
|---|---|---|---|
| concatenate sound representation with image without extra loss | audiovisual ResNet-101 + ResNet-50 · ADVANCE (audiovisual) | F-score: 72.85 vs. 72.71 | No clear effect |
| add sound event knowledge transfer losses (L_SQ, L_KL, L_E) to multimodal training | audiovisual ResNet-101 + ResNet-50 · ADVANCE (audiovisual) | F-score | Improved |
| add L_E (scene-to-event posterior transfer) rather than multitask L_SQ/L_KL | audiovisual ResNet-101 + ResNet-50 · ADVANCE (audiovisual) | F-score | Improved |
| add pre-trained weight initialization (image) (ablation) | audiovisual ResNet-101 + ResNet-50 · ADVANCE (image only) | F-score: 64.04 -> 72.85 | Improved |
| add pre-trained weight initialization (sound) (ablation) | audiovisual ResNet-101 + ResNet-50 · ADVANCE (sound only) | F-score: 21.32 -> 28.99 | Improved |
| add L_SQ|Nv+a to image-only network | audiovisual ResNet-101 + ResNet-50 · ADVANCE (image only) | F-score: 72.85 -> 73.98 | Improved |
| add L_E to image-only network | audiovisual ResNet-101 + ResNet-50 · ADVANCE (image only) | F-score: 72.85 -> 73.44 | Improved |
| add L_E to sound-only network | audiovisual ResNet-101 + ResNet-50 · ADVANCE (sound only) | F-score: 28.99 -> 29.66 | Improved |
| add L_KL|Nv+a to sound-only network | audiovisual ResNet-101 + ResNet-50 · ADVANCE (sound only) | F-score: 28.99 -> 28.51 | Hurt |
| add event relevance term beta*L_E2 to L_E1 (no Ls) (ablation) | audiovisual ResNet-101 + ResNet-50 · ADVANCE (no scene loss Ls) | F-score: 42.50 -> 48.65 | Improved |
| multitask L_KL|Nv+a instead of L_E when no scene loss | audiovisual ResNet-101 + ResNet-50 · ADVANCE (no scene loss Ls) | F-score: 4.46 vs. 48.65 | Hurt |
Closest papers by meaning
- Hear, Localize, and Reason: Spatially Aware Scene Understanding for Audio-visual LLMs (NeurIPS 2026 · similarity 0.84)
- Crab: A Unified Audio-Visual Scene Understanding Model with Explicit Cooperation (CVPR 2025 · similarity 0.84)
- Multiview Aerial Visual RECognition (MAVREC): Can Multi-view Improve Aerial Visual Perception? (CVPR 2024 · similarity 0.83)
- Multi-Modal Aerial-Ground Cross-View Place Recognition with Neural ODEs (CVPR 2025 · similarity 0.82)
- GeoViS: Geospatially Rewarded Visual Search for Remote Sensing Visual Grounding (CVPR 2026 · similarity 0.82)
- SkyMoE: A Vision-Language Foundation Model for Enhancing Geospatial Interpretation with Mixture of Experts (AAAI 2026 · similarity 0.82)
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.
