MG-LLaVA: Towards Multi-Granularity Visual Instruction Tuning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhao, Xiangyu, Li, Xiangtai, Duan, Haodong, Huang, Haian, Li, Yining, Chen, Kai, Yang, Hua
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929401212436480
author Zhao, Xiangyu
Li, Xiangtai
Duan, Haodong
Huang, Haian
Li, Yining
Chen, Kai
Yang, Hua
author_facet Zhao, Xiangyu
Li, Xiangtai
Duan, Haodong
Huang, Haian
Li, Yining
Chen, Kai
Yang, Hua
contents Multi-modal large language models (MLLMs) have made significant strides in various visual understanding tasks. However, the majority of these models are constrained to process low-resolution images, which limits their effectiveness in perception tasks that necessitate detailed visual information. In our study, we present MG-LLaVA, an innovative MLLM that enhances the model's visual processing capabilities by incorporating a multi-granularity vision flow, which includes low-resolution, high-resolution, and object-centric features. We propose the integration of an additional high-resolution visual encoder to capture fine-grained details, which are then fused with base visual features through a Conv-Gate fusion network. To further refine the model's object recognition abilities, we incorporate object-level features derived from bounding boxes identified by offline detectors. Being trained solely on publicly available multimodal data through instruction tuning, MG-LLaVA demonstrates exceptional perception skills. We instantiate MG-LLaVA with a wide variety of language encoders, ranging from 3.8B to 34B, to evaluate the model's performance comprehensively. Extensive evaluations across multiple benchmarks demonstrate that MG-LLaVA outperforms existing MLLMs of comparable parameter sizes, showcasing its remarkable efficacy. The code will be available at https://github.com/PhoenixZ810/MG-LLaVA.
format Preprint
id arxiv_https___arxiv_org_abs_2406_17770
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle MG-LLaVA: Towards Multi-Granularity Visual Instruction Tuning
Zhao, Xiangyu
Li, Xiangtai
Duan, Haodong
Huang, Haian
Li, Yining
Chen, Kai
Yang, Hua
Computer Vision and Pattern Recognition
Multi-modal large language models (MLLMs) have made significant strides in various visual understanding tasks. However, the majority of these models are constrained to process low-resolution images, which limits their effectiveness in perception tasks that necessitate detailed visual information. In our study, we present MG-LLaVA, an innovative MLLM that enhances the model's visual processing capabilities by incorporating a multi-granularity vision flow, which includes low-resolution, high-resolution, and object-centric features. We propose the integration of an additional high-resolution visual encoder to capture fine-grained details, which are then fused with base visual features through a Conv-Gate fusion network. To further refine the model's object recognition abilities, we incorporate object-level features derived from bounding boxes identified by offline detectors. Being trained solely on publicly available multimodal data through instruction tuning, MG-LLaVA demonstrates exceptional perception skills. We instantiate MG-LLaVA with a wide variety of language encoders, ranging from 3.8B to 34B, to evaluate the model's performance comprehensively. Extensive evaluations across multiple benchmarks demonstrate that MG-LLaVA outperforms existing MLLMs of comparable parameter sizes, showcasing its remarkable efficacy. The code will be available at https://github.com/PhoenixZ810/MG-LLaVA.
title MG-LLaVA: Towards Multi-Granularity Visual Instruction Tuning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2406.17770