Skip to main content
Cornell University
We gratefully acknowledge support from the Simons Foundation, member institutions, and all contributors. Donate
arxiv logo > cs > arXiv:2303.15343

Help | Advanced Search

arXiv logo
Cornell University Logo

quick links

  • Login
  • Help Pages
  • About

Computer Science > Computer Vision and Pattern Recognition

arXiv:2303.15343 (cs)
[Submitted on 27 Mar 2023 (v1), last revised 27 Sep 2023 (this version, v4)]

Title:Sigmoid Loss for Language Image Pre-Training

Authors:Xiaohua Zhai, Basil Mustafa, Alexander Kolesnikov, Lucas Beyer
View a PDF of the paper titled Sigmoid Loss for Language Image Pre-Training, by Xiaohua Zhai and 3 other authors
View PDF
Abstract:We propose a simple pairwise Sigmoid loss for Language-Image Pre-training (SigLIP). Unlike standard contrastive learning with softmax normalization, the sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities for normalization. The sigmoid loss simultaneously allows further scaling up the batch size, while also performing better at smaller batch sizes. Combined with Locked-image Tuning, with only four TPUv4 chips, we train a SigLiT model that achieves 84.5% ImageNet zero-shot accuracy in two days. The disentanglement of the batch size from the loss further allows us to study the impact of examples vs pairs and negative to positive ratio. Finally, we push the batch size to the extreme, up to one million, and find that the benefits of growing batch size quickly diminish, with a more reasonable batch size of 32k being sufficient. We release our models at this https URL and hope our research motivates further explorations in improving the quality and efficiency of language-image pre-training.
Comments: ICCV'23 Oral. arXiv v2: fix typo in pseudocode; v3: clarify t vs t' init; v4: add SigLIP Base, Large, Shape-Optimized 400M results. Models released at: this https URL. Xiaohua and Lucas contributed equally
Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
Cite as: arXiv:2303.15343 [cs.CV]
  (or arXiv:2303.15343v4 [cs.CV] for this version)
  https://j8ftcxlxtb.proxynodejs.usequeue.com/10.48550/arXiv.2303.15343
arXiv-issued DOI via DataCite

Submission history

From: Xiaohua Zhai [view email]
[v1] Mon, 27 Mar 2023 15:53:01 UTC (312 KB)
[v2] Thu, 30 Mar 2023 17:03:49 UTC (332 KB)
[v3] Thu, 4 May 2023 17:39:26 UTC (324 KB)
[v4] Wed, 27 Sep 2023 12:05:41 UTC (601 KB)
Full-text links:

Access Paper:

    View a PDF of the paper titled Sigmoid Loss for Language Image Pre-Training, by Xiaohua Zhai and 3 other authors
  • View PDF
  • Other Formats
license icon view license
Current browse context:
cs.CV
< prev   |   next >
new | recent | 2023-03
Change to browse by:
cs
cs.AI

References & Citations

  • NASA ADS
  • Google Scholar
  • Semantic Scholar
a export BibTeX citation Loading...

BibTeX formatted citation

×
Data provided by:

Bookmark

BibSonomy logo Reddit logo

Bibliographic and Citation Tools

Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)

Code, Data and Media Associated with this Article

alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)

Demos

Replicate (What is Replicate?)
Hugging Face Spaces (What is Spaces?)
TXYZ.AI (What is TXYZ.AI?)

Recommenders and Search Tools

Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
  • Author
  • Venue
  • Institution
  • Topic

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)
  • About
  • Help
  • contact arXivClick here to contact arXiv Contact
  • subscribe to arXiv mailingsClick here to subscribe Subscribe
  • Copyright
  • Privacy Policy
  • Web Accessibility Assistance
  • arXiv Operational Status
    Get status notifications via email or slack