Research map / Vision-language and multimodal / Cross-modal learning
ImageBindDC: Compressing Multi-modal Data with ImageBind-based Condensation
Yue Min, Shaobo Wang, Jiaze Li, Tianle Niu, Junxin Fan, Yongliang Miao, Lijin Yang, Linfeng Zhang
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 cross-modal and joint-modal alignment to uni-modal CF matching (ablation) | ImageBindDC · AVE (10 DPC) | accuracy: 70.34 -> 73.67 | Improved |
| replace MMD (DM) with characteristic function discrepancy (ablation) | DM · AVE/VGGS uni-modal audio (1 DPC) | accuracy: 27.87 -> 32.33 | Improved |
| replace MMD (DM) with characteristic function discrepancy (ablation) | DM · Video-Audio (10 DPC) | accuracy: 69.26 -> 70.34 | Improved |
| condense in ImageBind shared space with three CF objectives vs AVDD | AVDD · VGGS-10K (10 DPC, ConvNet condensation) | accuracy: 48.08 -> 55.23 | Improved |
| condense in ImageBind shared space with three CF objectives vs AVDD | AVDD · AVE (1 DPC, ImageBind-guided) | accuracy: 67.32 -> 70.10 | Improved |
| ImageBindDC vs AVDD on NYU-v2 at 10 DPC | AVDD · NYU-v2 (10 DPC) | accuracy: 98.62 -> 98.73 | No clear effect |
| ImageBindDC vs AVDD on NYU-v2 at 1 DPC | AVDD · NYU-v2 (1 DPC) | accuracy: 72.22 -> 80.43 | Improved |
| use ImageBindDC for condensation instead of DM/AVDD | DM · VGGS-10K (1 DPC) | condensation time (s): 140.30 -> 57.46 | Improved |
| ImageBindDC vs DM on ConvNet evaluation (AVE 1 DPC, ImageBind-guided condensation) | DM · AVE ConvNet (1 DPC) | accuracy: 7.10 -> 12.69 | Improved |
| ImageBindDC cross-architecture ConvNet at 10 DPC | AVDD · AVE ConvNet (10 DPC) | accuracy: 12.44 -> 16.74 | Mixed |
| ImageBindDC audio-text retrieval vs whole data | ImageBindDC · Clotho (20 DPC) | A2T R@10: 0.1995 vs 0.2364 whole | No clear effect |
Closest papers by meaning
- AlignMamba: Enhancing Multimodal Mamba with Local and Global Cross-modal Alignment (CVPR 2025 · similarity 0.87)
- UNIALIGN: Scaling Multimodal Alignment within One Unified Model (CVPR 2025 · similarity 0.87)
- With Limited Data for Multimodal Alignment, Let the STRUCTURE Guide You (NeurIPS 2025 · similarity 0.86)
- Robust Noisy Correspondence Learning with Equivariant Similarity Consistency (CVPR 2024 · similarity 0.86)
- Cross the Gap: Exposing the Intra-modal Misalignment in CLIP via Modality Inversion (ICLR 2025 · similarity 0.85)
- CovMatch: Cross-Covariance Guided Multimodal Dataset Distillation with Trainable Text Encoder (NeurIPS 2025 · similarity 0.85)
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.
