LiM-YOLO: Less is More with Pyramid Level Shift and Normalized Auxiliary Branch for Ship Detection in Optical Remote Sensing Imagery

2025-12-10

AI Summary

Optical remote sensing images pose significant challenges for ship detection due to extremely small object scales and strong anisotropic shapes, leading to high miss rates and localization inaccuracies in general-purpose detectors. To address this, we propose a lightweight and efficient YOLO architecture specifically tailored for remote sensing ship detection. Our key contributions are: (1) a novel pyramid-level offset strategy that relocates detection heads to feature levels P2–P4, satisfying the Nyquist sampling criterion to preserve fine spatial details of slender ships; (2) a group-normalized linear projection convolutional block (GN-CBLinear) that enhances stability in high-resolution feature extraction and robustness under small-batch training; and (3) integration of remote sensing–specific data augmentation and post-processing techniques. Extensive experiments on SODA-A, DOTA-v1.5, FAIR1M-v2.0, and ShipRSImageNet-V1 demonstrate state-of-the-art performance—achieving higher accuracy, fewer parameters, and faster inference compared to existing methods.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Search and Optimization: Learning to Search
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Applying general-purpose object detectors to ship detection in satellite imagery presents significant challenges due to the extreme scale disparity and morphological anisotropy of maritime targets. Standard architectures utilizing stride-32 (P5) layers often fail to resolve narrow vessels, resulting in spatial feature dilution. In this work, we propose LiM-YOLO, a specialized detector designed to resolve these domain-specific conflicts. Based on a statistical analysis of ship scales, we introduce a Pyramid Level Shift Strategy that reconfigures the detection head to P2-P4. This shift ensures compliance with Nyquist sampling criteria for small objects while eliminating the computational redundancy of deep layers. To further enhance training stability on high-resolution inputs, we incorporate a Group Normalized Convolutional Block for Linear Projection (GN-CBLinear), which mitigates gradient volatility in micro-batch settings. Validated on SODA-A, DOTA-v1.5, FAIR1M-v2.0, and ShipRSImageNet-V1, LiM-YOLO demonstrates superior detection accuracy and efficiency compared to state-of-the-art models. The code is available at https://github.com/egshkim/LiM-YOLO.

Problem

• Detects ships in satellite images with extreme scale variations
• Resolves spatial feature dilution for narrow vessels using pyramid shift
• Enhances training stability on high-resolution imagery via normalization

Innovation

• Pyramid Level Shift Strategy reconfigures detection heads to P2-P4
• Group Normalized Convolutional Block stabilizes training on high-resolution inputs
• Eliminates deep layer redundancy to enhance small ship detection