Nat Mini In1k 224

🚀 NAT (mini variant)

NAT-Mini is trained on ImageNet-1K at a resolution of 224x224. It offers a novel approach to image classification using Neighborhood Attention.

🚀 Quick Start

You can use the raw model for image classification. Check out the model hub to find fine - tuned versions for tasks that interest you.

Example

Here's how to classify an image from the COCO 2017 dataset into one of the 1,000 ImageNet classes using this model:

from transformers import AutoImageProcessor, NatForImageClassification
from PIL import Image
import requests

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

feature_extractor = AutoImageProcessor.from_pretrained("shi-labs/nat-mini-in1k-224")
model = NatForImageClassification.from_pretrained("shi-labs/nat-mini-in1k-224")

inputs = feature_extractor(images=image, return_tensors="pt")
outputs = model(**inputs)
logits = outputs.logits
# model predicts one of the 1000 ImageNet classes
predicted_class_idx = logits.argmax(-1).item()
print("Predicted class:", model.config.id2label[predicted_class_idx])

For more examples, refer to the documentation.

Requirements

Apart from transformers, this model requires the NATTEN package.

If you're on Linux, you can refer to shi-labs.com/natten for instructions on installing with pre - compiled binaries (just select your torch build to get the correct wheel URL).

Alternatively, you can use pip install natten to compile on your device, which may take up to a few minutes. Mac users only have the latter option (no pre - compiled binaries).

For more information, refer to NATTEN's GitHub.

✨ Features

• Neighborhood Attention: NAT is a hierarchical vision transformer based on Neighborhood Attention (NA). NA is a restricted self - attention pattern where each token's receptive field is limited to its nearest neighboring pixels. It is a sliding - window attention pattern, which is highly flexible and maintains translational equivariance.
• PyTorch Implementation: NA is implemented in PyTorch through its extension, NATTEN.

📚 Documentation

Model description

NAT is a hierarchical vision transformer based on Neighborhood Attention (NA). Neighborhood Attention is a restricted self - attention pattern in which each token's receptive field is limited to its nearest neighboring pixels. NA is a sliding - window attention pattern, and as a result, it is highly flexible and maintains translational equivariance.

NA is implemented in PyTorch implementations through its extension, NATTEN.

Source

BibTeX entry and citation info

@article{hassani2022neighborhood,
	title        = {Neighborhood Attention Transformer},
	author       = {Ali Hassani and Steven Walton and Jiachen Li and Shen Li and Humphrey Shi},
	year         = 2022,
	url          = {https://arxiv.org/abs/2204.07143},
	eprint       = {2204.07143},
	archiveprefix = {arXiv},
	primaryclass = {cs.CV}
}

📄 License

This project is licensed under the MIT license.

📦 Additional Information