Dragonfly: Multi-Resolution Zoom-In Encoding Enhances Vision-Language Models

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Thapa, Rahul, Chen, Kezhen, Covert, Ian, Chalamala, Rahul, Athiwaratkun, Ben, Song, Shuaiwen Leon, Zou, James
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913546463346688
author Thapa, Rahul
Chen, Kezhen
Covert, Ian
Chalamala, Rahul
Athiwaratkun, Ben
Song, Shuaiwen Leon
Zou, James
author_facet Thapa, Rahul
Chen, Kezhen
Covert, Ian
Chalamala, Rahul
Athiwaratkun, Ben
Song, Shuaiwen Leon
Zou, James
contents Recent advances in vision-language models (VLMs) have demonstrated the advantages of processing images at higher resolutions and utilizing multi-crop features to preserve native resolution details. However, despite these improvements, existing vision transformers (ViTs) still struggle to capture fine-grained details from less prominent objects, charts, and embedded text, limiting their effectiveness in certain tasks. In this paper, we extend recent high-resolution and multi-crop techniques by not only preserving the native resolution, but zooming in beyond it and extracting features from a large number of image sub-crops. This enhancement allows our model to better capture fine-grained details, overcoming the limitations of current ViTs. To manage the increased token count and computational complexity, we demonstrate that a simple mean-pooling aggregation over tokens is effective. Our model, Dragonfly, achieves competitive performance on general-domain tasks such as ScienceQA and AI2D, and excels in tasks requiring fine-grained image understanding, including TextVQA and ChartQA. Among models in the 7-8B parameter range, Dragonfly consistently ranks at the top across ten general-domain benchmarks, achieving the highest or second-highest scores in most cases, outperforming models that are significantly larger or trained on larger datasets. Our biomedical model, Dragonfly-Med, sets new benchmarks on several medical tasks, achieving 91.6% accuracy on SLAKE (compared to 84.8% for Med-Gemini), a 67.1% token F1 score on Path-VQA (compared to 62.7% for Med-PaLM M), and state-of-the-art results across the majority of image captioning tasks. Overall, our work highlights the persistent challenge of engineering visual representations with fixed-resolution ViTs, and proposes a simple yet effective solution to address this issue and boost performance in both general and specialized domains.
format Preprint
id arxiv_https___arxiv_org_abs_2406_00977
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Dragonfly: Multi-Resolution Zoom-In Encoding Enhances Vision-Language Models
Thapa, Rahul
Chen, Kezhen
Covert, Ian
Chalamala, Rahul
Athiwaratkun, Ben
Song, Shuaiwen Leon
Zou, James
Computer Vision and Pattern Recognition
Artificial Intelligence
Recent advances in vision-language models (VLMs) have demonstrated the advantages of processing images at higher resolutions and utilizing multi-crop features to preserve native resolution details. However, despite these improvements, existing vision transformers (ViTs) still struggle to capture fine-grained details from less prominent objects, charts, and embedded text, limiting their effectiveness in certain tasks. In this paper, we extend recent high-resolution and multi-crop techniques by not only preserving the native resolution, but zooming in beyond it and extracting features from a large number of image sub-crops. This enhancement allows our model to better capture fine-grained details, overcoming the limitations of current ViTs. To manage the increased token count and computational complexity, we demonstrate that a simple mean-pooling aggregation over tokens is effective. Our model, Dragonfly, achieves competitive performance on general-domain tasks such as ScienceQA and AI2D, and excels in tasks requiring fine-grained image understanding, including TextVQA and ChartQA. Among models in the 7-8B parameter range, Dragonfly consistently ranks at the top across ten general-domain benchmarks, achieving the highest or second-highest scores in most cases, outperforming models that are significantly larger or trained on larger datasets. Our biomedical model, Dragonfly-Med, sets new benchmarks on several medical tasks, achieving 91.6% accuracy on SLAKE (compared to 84.8% for Med-Gemini), a 67.1% token F1 score on Path-VQA (compared to 62.7% for Med-PaLM M), and state-of-the-art results across the majority of image captioning tasks. Overall, our work highlights the persistent challenge of engineering visual representations with fixed-resolution ViTs, and proposes a simple yet effective solution to address this issue and boost performance in both general and specialized domains.
title Dragonfly: Multi-Resolution Zoom-In Encoding Enhances Vision-Language Models
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2406.00977