Fig 1.
YOLOv7-tiny network structure.
Fig 2.
FB-YOLOv7 network structure.
Fig 3.
E-BRA structure.
Fig 4.
Transformer, BRA, E-BRA feature extraction area.
Table 1.
The parameters of the improved model are compared in task performance.
Fig 5.
AFPN structure.
Fig 6.
(a) Box loss comparison of loss function. (b) Cls loss comparison of loss function.
Table 2.
The overall and classification accuracy of YOLOv7-tiny and FB-YOLOv7 on two datasets.
Table 3.
Ablation experimental data.
Fig 7.
(a) Comparison of mainstream models in Helmet detection (b) Comparison of mainstream models in Daylight-v1.
Table 4.
Experimental data from the mainstream model on each of the two datasets.
Fig 8.
Visualization of detection results for different models.