EarthBridge: A Solution for 4th Multi-modal Aerial View Image Challenge Translation Track

2026-03-06

AI Summary

This work addresses the challenge of cross-modal translation among electro-optical (EO), infrared (IR), and synthetic aperture radar (SAR) aerial imagery, which arises from their distinct electromagnetic and geometric characteristics. To this end, the authors propose EarthBridge, a novel framework that integrates a Diffusion Bridge Implicit Model (DBIM) with Contrastive Unpaired Translation (CUT). EarthBridge innovatively employs a non-Markovian bridge process to enable high-quality deterministic sampling and introduces a channel-concatenated UNet denoiser alongside a tailored “warm-start noise” initialization mechanism to effectively mitigate ambiguity in cross-modal mapping. Evaluated on all four tasks of the MAVIC-T challenge, EarthBridge achieves a second-place overall score of 0.38, demonstrating significant improvements in spatial detail and spectral fidelity of the generated images.

Technology Category

• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

Cross-modal image-to-image translation among Electro-Optical (EO), Infrared (IR), and Synthetic Aperture Radar (SAR) sensors is essential for comprehensive multi-modal aerial-view analysis. However, translating between these modalities is notoriously difficult due to their distinct electromagnetic signatures and geometric characteristics. This paper presents \textbf{EarthBridge}, a high-fidelity translation framework developed for the 4th Multi-modal Aerial View Image Challenge -- Translation (MAVIC-T). We explore two distinct methodologies: \textbf{Diffusion Bridge Implicit Models (DBIM)}, which we generalize using non-Markovian bridge processes for high-quality deterministic sampling, and \textbf{Contrastive Unpaired Translation (CUT)}, which utilizes contrastive learning for structural consistency. Our EarthBridge framework employs a channel-concatenated UNet denoiser trained with Karras-weighted bridge scalings and a specialized"booting noise"initialization to handle the inherent ambiguity in cross-modal mappings. We evaluate these methods across all four challenge tasks (SAR$\rightarrow$EO, SAR$\rightarrow$RGB, SAR$\rightarrow$IR, RGB$\rightarrow$IR), achieving superior spatial detail and spectral accuracy. Our solution achieved a composite score of 0.38, securing the second position on the MAVIC-T leaderboard. Code is available at https://github.com/Bili-Sakura/EarthBridge-Preview.

Problem

• cross-modal image translation
• multi-modal aerial imagery
• EO-IR-SAR translation
• image-to-image translation
• aerial view analysis

Innovation

• Diffusion Bridge Implicit Models
• Contrastive Unpaired Translation
• Non-Markovian Bridge Processes
• Cross-modal Image Translation
• Bootstrapped Noise Initialization