Lightweight RGB-T Tracking with Mobile Vision Transformers

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Falaki, Mahdi, Amer, Maria A.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915769166594048
author Falaki, Mahdi
Amer, Maria A.
author_facet Falaki, Mahdi
Amer, Maria A.
contents Single-modality tracking (RGB-only) struggles under low illumination, weather, and occlusion. Multimodal tracking addresses this by combining complementary cues. While Vision Transformer-based trackers achieve strong accuracy, they are often too large for real-time. We propose a lightweight RGB-T tracker built on MobileViT with a progressive fusion framework that models intra- and inter-modal interactions using separable mixed attention. This design delivers compact, effective features for accurate localization, with under 4M parameters and real-time performance of 25.7 FPS on the CPU and 122 FPS on the GPU, supporting embedded and mobile platforms. To the best of our knowledge, this is the first MobileViT-based multimodal tracker. Model code and weights are available in the GitHub repository.
format Preprint
id arxiv_https___arxiv_org_abs_2506_19154
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Lightweight RGB-T Tracking with Mobile Vision Transformers
Falaki, Mahdi
Amer, Maria A.
Computer Vision and Pattern Recognition
Single-modality tracking (RGB-only) struggles under low illumination, weather, and occlusion. Multimodal tracking addresses this by combining complementary cues. While Vision Transformer-based trackers achieve strong accuracy, they are often too large for real-time. We propose a lightweight RGB-T tracker built on MobileViT with a progressive fusion framework that models intra- and inter-modal interactions using separable mixed attention. This design delivers compact, effective features for accurate localization, with under 4M parameters and real-time performance of 25.7 FPS on the CPU and 122 FPS on the GPU, supporting embedded and mobile platforms. To the best of our knowledge, this is the first MobileViT-based multimodal tracker. Model code and weights are available in the GitHub repository.
title Lightweight RGB-T Tracking with Mobile Vision Transformers
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2506.19154