PixDLM: A Dual-Path Multimodal Language Model for UAV Reasoning Segmentation

2026-04-16

AI Summary

This work addresses the challenges of semantic segmentation in unmanned aerial vehicle (UAV) remote sensing imagery, which suffers from oblique viewpoints, ultra-high resolution, and extreme scale variations. To this end, we formally define the UAV reasoning-aware segmentation task for the first time and decompose its semantic requirements into three reasoning dimensions: spatial, attribute, and scene. We introduce DRSeg, a large-scale benchmark dataset comprising 10,000 high-resolution aerial images annotated with Chain-of-Thought question-answer pairs, and propose PixDLM, a lightweight yet effective dual-path pixel-level multimodal language model as a unified baseline. Experimental results demonstrate that PixDLM establishes a strong performance baseline on DRSeg, effectively tackling the unique challenges of UAV reasoning-aware segmentation and providing a reliable foundation for future research in both data and methodology.

Technology Category

• Computer Vision: Segmentation
• Knowledge Representation and Reasoning: Common-Sense Reasoning
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Reasoning segmentation has recently expanded from ground-level scenes to remote-sensing imagery, yet UAV data poses distinct challenges, including oblique viewpoints, ultra-high resolutions, and extreme scale variations. To address these issues, we formally define the UAV Reasoning Segmentation task and organize its semantic requirements into three dimensions: Spatial, Attribute, and Scene-level reasoning. Based on this formulation, we construct DRSeg, a large-scale benchmark for UAV reasoning segmentation, containing 10k high-resolution aerial images paired with Chain-of-Thought QA supervision across all three reasoning types. As a benchmark companion, we introduce PixDLM, a simple yet effective pixel-level multimodal language model that serves as a unified baseline for this task. Experiments on DRSeg establish strong baseline results and highlight the unique challenges of UAV reasoning segmentation, providing a solid foundation for future research.

Problem

• UAV Reasoning Segmentation
• Oblique Viewpoints
• Ultra-high Resolution
• Scale Variations
• Multimodal Language Model

Innovation

• UAV Reasoning Segmentation
• Multimodal Language Model
• Chain-of-Thought Supervision
• Pixel-level Reasoning
• Remote Sensing Benchmark