Aimv2 Huge Patch14 448

🚀 Transformers Library for Image Feature Extraction

This library offers the AIMv2 family of vision models pre - trained with a multimodal autoregressive objective. These models are simple to train, scale effectively, and demonstrate excellent performance on various multimodal understanding benchmarks.

🚀 Quick Start

📦 Installation

The installation steps are not provided in the original document, so this section is skipped.

💻 Usage Examples

Basic Usage

The following examples demonstrate how to use the AIMv2 model for image feature extraction in both PyTorch and JAX.

PyTorch

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

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

processor = AutoImageProcessor.from_pretrained(
    "apple/aimv2-huge-patch14-448",
)
model = AutoModel.from_pretrained(
    "apple/aimv2-huge-patch14-448",
    trust_remote_code=True,
)

inputs = processor(images=image, return_tensors="pt")
outputs = model(**inputs)

JAX

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

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

processor = AutoImageProcessor.from_pretrained(
    "apple/aimv2-huge-patch14-448",
)
model = FlaxAutoModel.from_pretrained(
    "apple/aimv2-huge-patch14-448",
    trust_remote_code=True,
)

inputs = processor(images=image, return_tensors="jax")
outputs = model(**inputs)

✨ Features

• Outperforms OAI CLIP and SigLIP on the majority of multimodal understanding benchmarks.
• Outperforms DINOv2 on open - vocabulary object detection and referring expression comprehension.
• Exhibits strong recognition performance with AIMv2 - 3B achieving 89.5% on ImageNet using a frozen trunk.

📚 Documentation

Model Information

Performance Metrics

The aimv2 - huge - patch14 - 448 model has been tested on multiple datasets for classification tasks, and the accuracy metrics are as follows:

Introduction

[AIMv2 Paper] [BibTeX]

We introduce the AIMv2 family of vision models pre - trained with a multimodal autoregressive objective. AIMv2 pre - training is simple and straightforward to train and scale effectively.

📄 License

The library is licensed under the apple - amlr license.

📄 Citation

If you find our work useful, please consider citing us as:

@misc{fini2024multimodalautoregressivepretraininglarge,
  author      = {Fini, Enrico and Shukor, Mustafa and Li, Xiujun and Dufter, Philipp and Klein, Michal and Haldimann, David and Aitharaju, Sai and da Costa, Victor Guilherme Turrisi and Béthune, Louis and Gan, Zhe and Toshev, Alexander T and Eichner, Marcin and Nabi, Moin and Yang, Yinfei and Susskind, Joshua M. and El - Nouby, Alaaeldin},
  url         = {https://arxiv.org/abs/2411.14402},
  eprint      = {2411.14402},
  eprintclass = {cs.CV},
  eprinttype  = {arXiv},
  title       = {Multimodal Autoregressive Pre - training of Large Vision Encoders},
  year        = {2024},
}