Aimv2 Large Patch14 224 Distilled

🚀 AIMv2 Vision Models

This repository introduces the AIMv2 family of vision models. These models are pre - trained with a multimodal autoregressive objective, offering simplicity in pre - training, easy scalability, and outstanding performance in various multimodal and vision - related tasks.

🚀 Quick Start

The AIMv2 models can be used for image feature extraction. You can use them with either PyTorch or JAX.

✨ Features

• Multimodal Benchmark Performance: Outperforms OAI CLIP and SigLIP on the majority of multimodal understanding benchmarks.
• Object Detection and Comprehension: Outperforms DINOv2 on open - vocabulary object detection and referring expression comprehension.
• Strong Recognition Ability: AIMv2 - 3B achieves 89.5% on ImageNet using a frozen trunk.

📦 Installation

The installation process is mainly related to the transformers library. You can install it via pip:

pip install transformers

💻 Usage Examples

Basic Usage

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-large-patch14-224-distilled",
)
model = AutoModel.from_pretrained(
    "apple/aimv2-large-patch14-224-distilled",
    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-large-patch14-224-distilled",
)
model = FlaxAutoModel.from_pretrained(
    "apple/aimv2-large-patch14-224-distilled",
    trust_remote_code=True,
)

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

📚 Documentation

[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 project uses 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},
}

Additional Information