.png){kind=icon}
Return
Low-light Object Detection Algorithm Based on Improved YOLOv11s
.png){kind=icon}
PRE
AI
DOI：.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Object detection in low-light scenarios remains challenging due to low contrast, severe noise, and color distortion, which significantly degrade feature representation and localization reliability. To tackle these issues, this paper proposes an improved YOLOv11s-based detector, termed SSD-YOLOs, for robust low-light object detection. First, we design a DConv-C3k2 module to improve noise-robust feature extraction in low signal-to-noise ratio (SNR) environments. Second, we introduce SPD-Conv into the model to enhance the capture of fine-grained details, such as edges and textures. Third, an SE (Squeeze-and-Excitation) block is embedded in the detection head to recalibrate channel-wise responses and adaptively improve discriminative learning. Furthermore, we propose PA_Inception, which integrates CNN-based local feature extraction with the global and local modeling capacity of Patch Aware, enabling multi-scale and multi-path feature enhancement. Experimental results on the ExDark dataset demonstrate that SSD-YOLOs achieves a 2.93% mAP@0.5 and 2.03% mAP@0.5:0.95 improvement over YOLOv11s. Validation on the DarkLight Images dataset further confirms the model's generalizability, with mAP@0.5 and mAP@0.5:0.95 improvements of 2.18% and 1.22%, respectively. These results validate the effectiveness of our algorithm for low-light object detection.
Keywords:
Low-light object detection
YOLOv11s
Attention mechanism
Dynamic convolution
Space-to-depth convolution
Patch Aware
Journal
E
IF:
0.6
Papers:
205
Citations:
0
Organization
Cited Papers
No cited papers available