LCW-YOLO: A Lightweight Multi-Scale Object Detection Method Based on YOLOv11 and Its Performance Evaluation in...

Area of Science:

• Computer Vision
• Machine Learning

Background:

• Object detection is crucial for computer vision.
• Detecting small objects in complex backgrounds is difficult due to feature loss and inefficiency.

Purpose of the Study:

• Introduce LCW-YOLO, a lightweight object detection framework.
• Improve accuracy and inference speed for detecting small objects in challenging conditions.

Main Methods:

• Utilize Wavelet Pooling for multi-frequency reconstruction to preserve details and reduce noise.
• Employ a CGBlock-enhanced C3K2 structure for efficient fusion of shallow and deep features.
• Incorporate an improved LDHead for enhanced classification and localization.

Main Results:

• LCW-YOLO outperforms existing detectors in accuracy and speed on public datasets.
• Demonstrates significant advantages in small-object, sparse, and cluttered environments.
• Achieves stronger representations under complex conditions with efficient feature fusion.

Conclusions:

• The proposed method advances resource-efficient detection models.
• LCW-YOLO is suitable for safety-critical and real-time applications.
• Multi-frequency feature preservation and efficient fusion are key to robust object detection.