SigLIP
Improved vision-language model by Google using sigmoid loss for contrastive learning.
About
SigLIP by Google is a vision-language model that replaces the softmax contrastive loss of CLIP with a sigmoid loss computed on each image-text pair, which scales better and improves zero-shot performance. It produces strong image and text embeddings for zero-shot classification and retrieval and is available through Hugging Face. It is trained with the big_vision JAX codebase. Released under the Apache 2.0 license.
Reviews (0)
Leave a Review
No reviews yet. Be the first to review!
Details
- Price
- Free
- Platform
- Local/Desktop
- Difficulty
- Intermediate (3/5)
- License
- Apache-2.0
- Minimum VRAM
- 4 GB
- Added
- Apr 3, 2026
Related Tools
Simple and effective multi-object tracking using every detection box.
Monocular depth estimation model producing detailed depth maps from single images.
End-to-end object detection with transformers by Meta, eliminating hand-designed components.
Self-supervised vision transformer by Meta producing universal visual features.
Unified vision foundation model by Microsoft for captioning, detection, and segmentation.
Robust multi-object tracking combining motion and appearance cues.