README.md

---
pipeline_tag: image-classification
tags:
  - vision
inference: false
widget:
  - src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/cat-dog-music.png
    example_title: Cat & Dog
---

# Category Search from External Databases (CaSED)

Disclaimer: The model card is taken and modified from the official repository, which can be found [here](https://github.com/altndrr/vic). The paper can be found [here](https://arxiv.org/abs/2306.00917).

## Intended uses & limitations

You can use the model for vocabulary-free image classification, i.e. classification with CLIP-like models without a pre-defined list of class names.

## How to use

Here is how to use this model:

```python
import requests
from PIL import Image
from transformers import AutoModel, CLIPProcessor

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

# load the model and the processor
model = AutoModel.from_pretrained("altndrr/cased", trust_remote_code=True)
processor = CLIPProcessor.from_pretrained("openai/clip-vit-large-patch14")

# get the model outputs
images = processor(images=[image], return_tensors="pt", padding=True)
outputs = model(images, alpha=0.7)
labels, scores = outputs["vocabularies"][0], outputs["scores"][0]

# print the top 5 most likely labels for the image
values, indices = scores.sort(dim=-1, descending=True)
print("\nTop predictions:\n")
for value, index in zip(values, indices):
    print(f"{labels[index]:>16s}: {100 * value.item():.2f}%")
```

The model depends on some libraries you have to install manually before execution:

```bash
pip install torch faiss-cpu flair inflect nltk pyarrow transformers
```

## Citation

```latex
@article{conti2023vocabularyfree,
      title={Vocabulary-free Image Classification},
      author={Alessandro Conti and Enrico Fini and Massimiliano Mancini and Paolo Rota and Yiming Wang and Elisa Ricci},
      year={2023},
      journal={NeurIPS},
}
```