Title: Paper page - SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic
  Understanding, Localization, and Dense Features
Description: Join the discussion on this paper page
Keywords: No keywords
Text content: Paper page - SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic   Understanding, Localization, and Dense Features         Hugging Face        					Models  					Datasets  					Spaces  					Posts  					Docs  					Enterprise  Pricing 			       Log In 				 Sign Up 					    			Papers 		 arxiv:2502.14786   SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic   Understanding, Localization, and Dense Features Published on Feb 20 ·  Submitted by  							akhaliq  on Feb 21 #2 Paper of the day   		Upvote  		87                   +79  Authors:  Michael Tschannen   					,  Alexey Gritsenko   					, Xiao Wang 						 					, Muhammad Ferjad Naeem 						 					,  Ibrahim Alabdulmohsin   					, Nikhil Parthasarathy 						 					,  Talfan Evans   					,  Lucas Beyer   					, Ye Xia 						 					,  Basil Mustafa   					,  Olivier Hénaff   					,  Jeremiah Harmsen   					, Andreas Steiner 						 					,  Xiaohua Zhai      Abstract We introduce SigLIP 2, a family of new multilingual vision-language encoders that build on the success of the original SigLIP. In this second iteration, we extend the original image-text training objective with several prior, independently developed techniques into a unified recipe -- this includes captioning-based pretraining, self-supervised losses (self-distillation, masked prediction) and online data curation. With these changes, SigLIP 2 models outperform their SigLIP counterparts at all model scales in core capabilities, including zero-shot classification, image-text retrieval, and transfer performance when extracting visual representations for Vision-Language Models (VLMs). Furthermore, the new training recipe leads to significant improvements on localization and dense prediction tasks. We also train variants which support multiple resolutions and preserve the input's native aspect ratio. Finally, we train on a more diverse data-mixture that includes de-biasing techniques, leading to much better multilingual understanding and improved fairness. To allow users to trade off inference cost with performance, we release model checkpoints at four sizes: ViT-B (86M), L (303M), So400m (400M), and g (1B). View arXiv page View PDF  			Add to collection 		 Community 	      akhaliq  Paper submitter 1 day ago         https://github.com/google-research/big_vision/blob/main/big_vision/configs/proj/image_text/README_siglip2.md  See translation    + 							  Reply       TronEasy  about 22 hours ago   This comment has been hidden   + 							         AntonioNocerino99  about 16 hours ago         It is truly magnificent! Thank you for sharing this work.  See translation    + 							  Reply       librarian-bot  about 10 hours ago         This is an automated message from the Librarian Bot. I found the following papers similar to this paper.  The following papers were recommended by the Semantic Scholar API   FLAVARS: A Multimodal Foundational Language and Vision Alignment Model for Remote Sensing (2025) Generate, Transduct, Adapt: Iterative Transduction with VLMs (2025) Efficient Few-Shot Continual Learning in Vision-Language Models (2025) MoVE-KD: Knowledge Distillation for VLMs with Mixture of Visual Encoders (2025) Centurio: On Drivers of Multilingual Ability of Large Vision-Language Model (2025) Unifying Specialized Visual Encoders for Video Language Models (2025) SCOT: Self-Supervised Contrastive Pretraining For Zero-Shot Compositional Retrieval (2025)   Please give a thumbs up to this comment if you found it helpful!  If you want recommendations for any Paper on Hugging Face checkout this Space  You can directly ask Librarian Bot for paper recommendations by tagging it in a comment:  @librarian-bot  recommend  See translation    + 							  Reply       franchesoni  about 1 hour ago         Not compared against dinov2reg or other SSL backbones for vision tasks...  See translation    + 							  Reply   EditPreview     Upload images, audio, and videos by dragging in the text input, pasting, or clicking here. 			 Tap or paste here to upload images     				Comment 			 · Sign up or 					log in to comment 				    		Upvote  		87                           +75   				Models citing this paper 				79  google/siglip2-base-patch16-224   			Zero-Shot Image Classification 			•  Updated 					1 day ago •   					364 				 	 				•   					15  				  google/siglip2-giant-opt-patch16-384   			Zero-Shot Image Classification 			•  Updated 					1 day ago •   					203 				 	 				•   					6  				  google/siglip2-large-patch16-512   			Zero-Shot Image Classification 			•  Updated 					1 day ago •   					58 				 	 				•   					5  				  google/siglip2-so400m-patch16-naflex   			Zero-Shot Image Classification 			•  Updated 					1 day ago •   					234 				 	 				•   					5  				  Browse 79 models citing this paper 					  				Datasets citing this paper 				0 No dataset linking this paper Cite arxiv.org/abs/2502.14786 in a dataset README.md to link it from this page. 				  				Spaces citing this paper 				2 🚀 google/zero-shot-sg1-sg2🤔 adorkin/siglip2-clothes   				Collections including this paper 				9  SigLIP2  					Collection 				  36 items •  Updated 					1 day ago •   					34  Papers  					Collection 				  348 items •  Updated 					about 6 hours ago •   					9  Papers to Read  					Collection 				  25 items •  Updated 					about 17 hours ago •   					1  readings  					Collection 				  229 items •  Updated 					about 22 hours ago   Browse 9 collections that include this paper 						     			System theme 		  Company TOS Privacy About Jobs  Website Models Datasets Spaces Pricing Docs