Research map / Vision-language and multimodal / Cross-modal learning
ImageBind: One Embedding Space To Bind Them All
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 |
|---|---|---|---|
| Align each non-image modality to image embeddings, without paired audio-text examples | ImageBind · Clotho | Recall@1: 3.0 -> 6.0 | Helped |
| Align each non-image modality to image embeddings, without paired audio-text examples | ImageBind · Clotho | Recall@10: 17.5 -> 28.4 | Helped |
| Combine audio and video embeddings for retrieval | ImageBind · MSR-VTT 1K-A | Recall@1: 6.8 -> 36.8 | Helped |
| Use a fixed rather than learnable loss temperature (ablation) | ImageBind · ESC fold-1 | top-1 accuracy: 54.8 -> 56.7 | Helped |
| Set loss temperature to 0.07 rather than 0.05 (ablation) | ImageBind · ESC fold-1 | top-1 accuracy: 56.7 -> 52.4 | Hurt |
| Set loss temperature to 1.0 rather than 0.05 (ablation) | ImageBind · ESC fold-1 | top-1 accuracy: 56.7 -> 24.3 | Hurt |
| Replace the linear projection head with an MLP (ablation) | ImageBind · ESC fold-1 | top-1 accuracy: 56.7 -> 51.0 | Hurt |
| Train for more epochs (ablation) | ImageBind · ESC fold-1 | top-1 accuracy: 56.7 -> 62.9 | Helped |
| Use strong rather than basic image augmentation (ablation) | ImageBind · SUN-D | top-1 accuracy: 25.4 -> 26.7 | Helped |
| Use strong rather than basic image augmentation (ablation) | ImageBind · ESC fold-1 | top-1 accuracy: 56.7 -> 22.6 | Hurt |
| Keep image and depth crops spatially aligned (ablation) | ImageBind · SUN-D | top-1 accuracy: 16.0 -> 26.7 | Helped |
| Add RandErase to depth augmentation (ablation) | ImageBind · SUN-D | top-1 accuracy: 24.2 -> 26.7 | Helped |
Closest papers by meaning
- UniBind: LLM-Augmented Unified and Balanced Representation Space to Bind Them All (CVPR 2024 · similarity 0.87)
- Cross the Gap: Exposing the Intra-modal Misalignment in CLIP via Modality Inversion (ICLR 2025 · similarity 0.86)
- Language-Driven Cross-Modal Classifier for Zero-Shot Multi-Label Image Recognition (ICML 2024 · similarity 0.86)
- LanguageBind: Extending Video-Language Pretraining to N-modality by Language-based Semantic Alignment (ICLR 2024 · similarity 0.86)
- Perception Encoder: The best visual embeddings are not at the output of the network (NeurIPS 2025 · similarity 0.86)
- Connect, Collapse, Corrupt: Learning Cross-Modal Tasks with Uni-Modal Data (ICLR 2024 · similarity 0.86)
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.