Research map / Vision-language and multimodal / CLIP and zero-shot
CF-IPT: Cross-Modal Fusion Interactive Prompt Tuning of Vision-Language Pre-Trained Model for Multisource Remote Sensing Data Classification
Jinheng Ji, Jiahui Qu, Wenqian Dong, Yunsong Li
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 |
|---|---|---|---|
| add spectral-spatial prompt matrices (vs all-one matrices) (ablation) | CF-IPT · Houston/MUUFL/Augsburg | OA (%): 96.74 -> 97.89 (Houston) | Improved |
| use fused image (vs channel concatenation) (ablation) | CF-IPT · Houston/MUUFL/Augsburg | OA (%): 93.28 -> 97.89 (Houston) | Improved |
| use learnable transposed-conv input adaptation encoder (vs bilinear upsampling) (ablation) | CF-IPT · Houston/MUUFL/Augsburg | OA (%): 92.49 -> 97.89 (Houston) | Improved |
| add text prompts to visual-only prompting (ablation) | CF-IPT · Houston/MUUFL/Augsburg | OA (%): 96.70 -> 97.89 (Houston) | Improved |
| add visual prompts to text-only prompting (ablation) | CF-IPT · Houston/MUUFL/Augsburg | OA (%): 96.22 -> 97.89 (Houston) | Improved |
| add visual-text prompt interaction (ablation) | CF-IPT · Houston/MUUFL/Augsburg | OA (%): 96.67 -> 97.89 (Houston) | Improved |
| use S2GPI Mamba transformation instead of a standard Transformer (ablation) | CF-IPT · Houston/MUUFL/Augsburg | OA (%): 96.85 -> 97.89 (Houston) | Improved |
| fine-tune CLIP with CF-IPT vs prior prompt methods on scene classification | CF-IPT · RSICD/PatternNet/RESISC45 | OA (%): 91.00 -> 92.80 (RSICD, vs MaPLe) | Improved |
| CF-IPT vs DiffCLIP with far fewer trainable parameters | CF-IPT · Houston | OA (%): 96.15 -> 97.89 OA; 10.581 -> 1.138 M params | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DiffCLIP — DiffCLIP: Few-shot Language-driven Multimodal Classifier (AAAI 2025) · better on 3 of 3 shared results
- MaPLe — MaPLe: Multi-Modal Prompt Learning (CVPR 2023) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MMM-RS: A Multi-modal, Multi-GSD, Multi-scene Remote Sensing Dataset and Benchmark for Text-to-Image Generation (NeurIPS 2024 · similarity 0.86)
- Meta-Learning Hyperparameters for Parameter Efficient Fine-Tuning (CVPR 2025 · similarity 0.86)
- Quality-Driven Curation of Remote Sensing Vision-Language Data via Learned Scoring Models (NeurIPS 2025 · similarity 0.85)
- Remote Sensing Vision-Language Foundation Models without Annotations via Ground Remote Alignment (ICLR 2024 · similarity 0.85)
- SkySense: A Multi-Modal Remote Sensing Foundation Model Towards Universal Interpretation for Earth Observation Imagery (CVPR 2024 · similarity 0.85)
- DiffCLIP: Few-shot Language-driven Multimodal Classifier (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.
