LMSFA-YOLO: A lightweight target detection network in Remote sensing images based on Multiscale feature fusion

Yuanbo Chu, Jiahao Wang, Lihong Ma, Chenxing WuJournal of King Saud University - Computer and Information Sciences 2025 · DOI

TL;DR

LMSFA-YOLO is a new, lightweight object detection framework designed specifically to find tiny targets in aerial remote sensing images. By rethinking how convolutions handle multiscale data and prioritizing high-resolution shallow features over deep, coarse ones, the authors achieved a 7.6% mAP boost over YOLOv5s while using 35.8% fewer parameters, all while maintaining real-time performance on edge hardware.

What Problem Does This Paper Address?

Detecting objects in remote sensing (aerial/satellite) imagery is notoriously difficult. Unlike standard datasets where objects are large and centered, aerial images are filled with "tiny targets"—cars, people, or small boats that may only be a few pixels wide.

Current state-of-the-art models face two major hurdles:

  1. Information Loss: Standard deep learning architectures use multiple downsampling layers. By the time the data reaches the deeper layers, the features of a tiny 10x10 pixel object are often "washed out" or lost entirely.
  2. Computational Cost: High-accuracy models are often too heavy to run on the edge devices (like drones or mobile terminals) that actually need them.

The authors aim to bridge this gap, creating a model that is both "smart" enough to see tiny objects and "lean" enough to run on a Jetson Orin Nano in real-time.

Methods

The authors introduced four key innovations to the YOLOv5s baseline:

  • LMSConv & LMSCSP: To save on computation, they designed a "Lightweight Multiscale Convolution." It splits input channels into two: one "cheap" branch that stays unchanged to preserve raw details, and one "main" branch that uses multiscale kernels (3x3 and 5x5) to capture context. This is integrated into a new Cross-Stage Partial (LMSCSP) module.
  • EMCASPP (Efficient Mixed Channel Attention SPP): They upgraded the standard Spatial Pyramid Pooling (SPPF) by adding Mixed Local Channel Attention (MLCA). This helps the model focus on important local features (the tiny targets) while ignoring redundant background noise.
  • High-Resolution Shallow Feature (HRSF) Layer: Recognizing that deep layers (P5) actually inhibit tiny object detection, they removed the deepest detection head and added a high-resolution shallow head (P2). This ensures the model looks at features with a smaller receptive field and higher spatial resolution.
  • ShapeIoU: Instead of using standard IoU (Intersection over Union) for the loss function, they used ShapeIoU. This metric specifically accounts for the shape and scale of bounding boxes, which is critical when the difference of a few pixels determines a successful detection for a tiny object.

Datasets

The model was evaluated on two challenging public datasets:

  • VisDrone: A diverse collection of drone-captured images where ~74.5% of objects are categorized as small (<32 pixels).
  • AI-TOD (Aerial Imagery-Tiny Object Detection): An even more extreme benchmark where 98.3% of objects are small, and 86% are "tiny" (<16 pixels).

Key Insights and Ideas

  • Deep isn't always better: One of the most interesting findings is that removing the deepest layer (P5) and focusing on the shallow P2 layer significantly improved accuracy. For tiny targets, the semantic abstraction of deep layers is actually a disadvantage because it destroys spatial precision.
  • The "Cheap" Branch Strategy: By keeping half the channels unprocessed in LMSConv, the model retains "fine-grained" features that standard convolutions might blur. This acts as a safety net for the tiny details required for detection.
  • Shape Matters: For tiny objects, a slight misalignment in the bounding box shape leads to a massive drop in IoU. ShapeIoU's focus on the aspect ratio and scale of the box proved much more effective than standard overlap-based metrics.

Key Results

The results show a rare "win-win" in model optimization:

  • Efficiency: LMSFA-YOLO uses only 4.52M parameters (a 35.8% reduction from YOLOv5s) and reduces model size from 14.1MB to 9.8MB.
  • Accuracy: On the VisDrone dataset, mAP0.5 improved by 7.0%. On the AI-TOD dataset, it improved by 7.6%.
  • Edge Performance: The model runs at ~32 FPS on a Jetson Orin Nano (4GB) with a low energy footprint (0.062 J/frame), making it truly viable for drone deployment.

Limitations and Future Work

While the model excels in dense and low-light scenarios, the authors noted it still struggles with extreme overexposure (strong light). Because the model relies heavily on shallow features, overexposed regions can "blind" the detector more easily than they would a deeper, more abstract model.

Future work will explore even more efficient architectures like Mamba or Swin Transformers and look into few-shot learning to help the model adapt to new environments with very little training data.

Summary updated

Abstract

In aerial target detection tasks, achieving a balance between high detection accuracy and low computational cost remains a key challenge. To address the aforementioned problems, this paper introduces the lightweight multiscale feature fusion and attention-YOLO (LMSFA-YOLO), which is lightweight and accurate for aerial tiny target recognition. Firstly, we propose a lightweight multiscale convolution (LMSConv) and a lightweight multiscale cross-stage partial (LMSCSP). These methods optimize convolutional computation cost and enhance multiscale information extraction, significantly reducing computational cost and parameters, while improving feature representation and fusion without sacrificing accuracy. Subsequently, the mixed local channel attention (MLCA) is combined to create an effective mixed channel attention spatial pyramid pooling (EMCASPP), aiming to simultaneously integrate local and channel space information to enhance the feature fusion ability of the model. To further improve the precision of feature extraction and preserve detailed information, a high-resolution shallow feature layer is applied. Finally, to increase the accuracy of bounding box regression, we introduce ShapeIoU to emphasize the scale and shape of the bounding box, replacing the original IoU. Experimental results demonstrate that LMSFA-YOLO surpasses the baseline YOLOv5s with 35.8% fewer parameters while improving the F1-score and mAP by 5.0% and 7.0% on the VisDrone dataset and by 6.1% and 7.6% on the AI-TOD dataset. Furthermore, it runs in real-time (> 30 FPS) on the Jetson Orin Nano. These results validate the effectiveness of LMSFA-YOLO in achieving high detection performance while reducing computation cost, making it well-suited for deployment on edge devices with limited computational resources.
BibTeX
@article{chu2025lmsfa,
  author = {Yuanbo Chu and Jiahao Wang and Lihong Ma and Chenxing Wu},
  title = {{LMSFA-YOLO: A lightweight target detection network in Remote sensing images based on Multiscale feature fusion}},
  journal = {Journal of King Saud University - Computer and Information Sciences},
  year = {2025},
  doi = {10.1007/s44443-025-00069-4},
}
Figure