Siglip Base Patch16 384

🚀 SigLIP (base-sized model)

A pre - trained multimodal model with a better loss function for zero - shot image classification and image - text retrieval.

🚀 Quick Start

SigLIP is a pre - trained model on WebLi at a resolution of 384x384. It combines the concept of CLIP with an improved loss function, enabling better performance in various multimodal tasks.

✨ Features

• Better Loss Function: The sigmoid loss in SigLIP operates on image - text pairs without the need for a global view of pairwise similarities for normalization, allowing for larger batch sizes and better performance at smaller batch sizes.
• Multimodal Capabilities: Suitable for tasks like zero - shot image classification and image - text retrieval.

📦 Installation

To use this model, you need to install the transformers library. You can install it via pip:

pip install transformers

You also need torch and Pillow for image processing. You can install them as follows:

pip install torch Pillow requests

💻 Usage Examples

Basic Usage

Here is how to use this model to perform zero - shot image classification:

from PIL import Image
import requests
from transformers import AutoProcessor, AutoModel
import torch

model = AutoModel.from_pretrained("google/siglip-base-patch16-384")
processor = AutoProcessor.from_pretrained("google/siglip-base-patch16-384")

url = "http://images.cocodataset.org/val2017/000000039769.jpg"
image = Image.open(requests.get(url, stream=True).raw)

texts = ["a photo of 2 cats", "a photo of 2 dogs"]
inputs = processor(text=texts, images=image, padding="max_length", return_tensors="pt")

with torch.no_grad():
    outputs = model(**inputs)

logits_per_image = outputs.logits_per_image
probs = torch.sigmoid(logits_per_image) # these are the probabilities
print(f"{probs[0][0]:.1%} that image 0 is '{texts[0]}'")

Advanced Usage

You can also leverage the pipeline API which abstracts away the complexity for the user:

from transformers import pipeline
from PIL import Image
import requests

# load pipe
image_classifier = pipeline(task="zero-shot-image-classification", model="google/siglip-base-patch16-384")

# load image
url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
image = Image.open(requests.get(url, stream=True).raw)

# inference
outputs = image_classifier(image, candidate_labels=["2 cats", "a plane", "a remote"])
outputs = [{"score": round(output["score"], 4), "label": output["label"] } for output in outputs]
print(outputs)

For more code examples, refer to the documentation.

📚 Documentation

Model Description

SigLIP is an enhanced version of CLIP, a multimodal model, with a better loss function. The sigmoid loss operates solely on image - text pairs and does not require a global view of the pairwise similarities for normalization. This allows further scaling up the batch size, while also performing better at smaller batch sizes.

A TLDR of SigLIP by one of the authors can be found here.

Intended Uses & Limitations

You can use the raw model for tasks like zero - shot image classification and image - text retrieval. Check the model hub for other versions on a task that interests you.

Training Procedure

Training Data

SigLIP is pre - trained on the English image - text pairs of the WebLI dataset (Chen et al., 2023).

Preprocessing

• Images: Resized/rescaled to the same resolution (384x384) and normalized across the RGB channels with mean (0.5, 0.5, 0.5) and standard deviation (0.5, 0.5, 0.5).
• Texts: Tokenized and padded to the same length (64 tokens).

Compute

The model was trained on 16 TPU - v4 chips for three days.

Evaluation Results

Evaluation of SigLIP compared to CLIP is shown below (taken from the paper).

BibTeX entry and citation info

@misc{zhai2023sigmoid,
      title={Sigmoid Loss for Language Image Pre-Training}, 
      author={Xiaohua Zhai and Basil Mustafa and Alexander Kolesnikov and Lucas Beyer},
      year={2023},
      eprint={2303.15343},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

📄 License

This model is licensed under the Apache 2.0 license.