Fig 1.
Schematic diagram of C3 and C2f module structure.
Fig 2.
Comparison of convolutional module structures.
(a) Ordinary convolutional module structure. (b) DSC module structure.
Fig 3.
(a) CBAM backbone structure. (b) Channel attention mechanism module. (c) Spatial attention mechanism module.
Fig 4.
Improved YOLOv8 network structure diagram.
Fig 5.
Comparison of the effect of different image expansion methods.
Fig 6.
(a) Image import. (b) Image segmentation annotation.
Fig 7.
Image annotation results.
Table 1.
Results of ablation experiments.
Table 2.
Comparison of detection results of different models.
Fig 8.
Comparison of partial detection effect of different models (case 1).
(a) original image (b) YOLOv3 model (c) YOLOv5 model (d) YOLOv7 model (e) YOLOv8 model (f) this paper.
Fig 9.
Comparison of partial detection effect of different models (case 2).
(a) original image (b) YOLOv3 model (c) YOLOv5 model (d) YOLOv7 model (e) YOLOv8 model (f) this paper.
Fig 10.
Comparison of partial detection effect of different models (case 3).
(a) original image (b) YOLOv3 model (c) YOLOv5 model (d) YOLOv7 model (e) YOLOv8 model (f) this paper.
Fig 11.
Confusion matrix of different models.
(a) YOLOv3 model (b) YOLOv5 model (c) YOLOv7 model (d) YOLOv8 model (e) this paper.
Fig 12.
Segmentation results of coal gangue image on working face belt conveyor.
(a) Underground gangue images (captured on belt conveyor) (b) Improved yolov8 model segmentation results.
Fig 13.
Heat map rendering of coal and gangue location.
Fig 14.
Recognition effect in case of mixed coal and gangue accumulation.
(a) original image (b) recognition effect.