Fig 1.
Framework of LightweightSRNet.
Fig 2.
Overall architecture of the improved YOLOv8 network.
Fig 3.
Structure of the HG-MHA.
Fig 4.
Comparison of SC-BiFPN with other fusion methods.
Fig 5.
Schematic Diagram of the Ghost Bottleneck Structure.
Fig 6.
Structural Workflow of the C2f-Ghost-Sobel Module.
Fig 7.
Target Distribution in the HIT-UAV Dataset.
Fig 8.
Comparison of Images Before and After Super-Resolution Reconstruction.
(a) Before Reconstruction; (b) After Reconstruction; (c) GT.
Table 1.
Comparison of detection performance using different RS methods.
Table 2.
Comparison of different attention mechanisms in infrared object detection.
Table 3.
Results of ablation experiments.
Fig 9.
Comparison of Training Results.
Fig 10.
Attention Distribution Visualization.
(a) Without Attention Mechanism; (b) Original Infrared Image and Targets; (c) With Attention Mechanism.
Table 4.
mAP (%) of each dataset under different object detection algorithms.
Fig 11.
Detection results on the HIT-UAV dataset using different algorithms.
From (a) to (g), the results correspond to Faster R-CNN, SSD, YOLOv5s, YOLOv7, YOLO-IR-Free, YOLO-DeepOC-IR, and ours, respectively.
Fig 12.
Detection results on the FLIR dataset using different algorithms.
From (a) to (g), the results correspond to Faster R-CNN, SSD, YOLOv5s, YOLOv7, YOLO-IR-Free, YOLO-DeepOC-IR, and ours, respectively.