Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

ImageBind: One Embedding Space To Bind Them All

About

We present ImageBind, an approach to learn a joint embedding across six different modalities - images, text, audio, depth, thermal, and IMU data. We show that all combinations of paired data are not necessary to train such a joint embedding, and only image-paired data is sufficient to bind the modalities together. ImageBind can leverage recent large scale vision-language models, and extends their zero-shot capabilities to new modalities just by using their natural pairing with images. It enables novel emergent applications 'out-of-the-box' including cross-modal retrieval, composing modalities with arithmetic, cross-modal detection and generation. The emergent capabilities improve with the strength of the image encoder and we set a new state-of-the-art on emergent zero-shot recognition tasks across modalities, outperforming specialist supervised models. Finally, we show strong few-shot recognition results outperforming prior work, and that ImageBind serves as a new way to evaluate vision models for visual and non-visual tasks.

Rohit Girdhar, Alaaeldin El-Nouby, Zhuang Liu, Mannat Singh, Kalyan Vasudev Alwala, Armand Joulin, Ishan Misra• 2023

Related benchmarks

TaskDatasetResultRank
Image ClassificationImageNet-1K
Top-1 Acc77.7
836
Language UnderstandingMMLU
Accuracy43.6
756
Text-to-Image RetrievalFlickr30k (test)
Recall@174.9
423
Action RecognitionUCF101--
365
Text-to-Video RetrievalDiDeMo
R@10.36
360
Audio ClassificationESC-50
Accuracy66.9
325
Text-to-Video RetrievalMSR-VTT
Recall@136.8
313
Text-to-Video RetrievalMSR-VTT (test)
R@136.8
234
Text-to-Video RetrievalMSVD
R@147.9
218
Text-to-Video RetrievalMSVD (test)
R@139.3
204
Showing 10 of 208 rows
...

Other info

Code

Follow for update