Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Fig 1.

YOLOv8n network architecture.

Backbone, which is responsible for feature extraction; Neck, which aggregates and fuses multi-scale features; and Head, which performs the final prediction.

More »

Fig 1 Expand

Fig 2.

Improved network model.

Compared with the baseline network, the backbone and neck are enhanced by incorporating GSConv and VoVGSCSP modules, while the C2f-RFAConv module is introduced to improve feature representation and information fusion efficiency.

More »

Fig 2 Expand

Fig 3.

RFAConv structure.

RFAConv integrates receptive-field spatial features to expand contextual perception and group convolution to improve computational efficiency and reduce parameter redundancy.

More »

Fig 3 Expand

Fig 4.

CRFAConv structure.

The CRFAConv module integrates a CBS block with RFAConv, enabling effective feature extraction through receptive-field spatial features and efficient convolution operations.

More »

Fig 4 Expand

Fig 5.

C2f structure.

C2f is a feature fusion module employed in the neck of YOLOv8 to enhance information flow and multi-scale feature representation.

More »

Fig 5 Expand

Fig 6.

C2f-RFAConv structure.

The original C2f module is enhanced by incorporating CRFAConv, aiming to strengthen feature representation while maintaining computational efficiency.

More »

Fig 6 Expand

Fig 7.

GSConv structure.

GSConv consists of a depthwise convolution (DWConv) for lightweight spatial feature extraction and a channel shuffle operation to promote information exchange across channel groups.

More »

Fig 7 Expand

Fig 8.

VOVGSCSP structure.

The VoVGSCSP module incorporates GSConv to achieve efficient feature extraction.

More »

Fig 8 Expand

Fig 9.

YOLOv8n detection head structure.

The detection head adopts a decoupled design to separately perform bounding box regression and classification on multi-scale feature maps.

More »

Fig 9 Expand

Fig 10.

SCConv structure.

The SCConv module is composed of a Spatial Reconstruction Unit (SRU) and a Channel Reconstruction Unit (CRU), which collaboratively enhance spatial and channel-wise feature representation.

More »

Fig 10 Expand

Fig 11.

SCConv Head structure.

SCConv is embedded into the coupled detection head to refine shared features for both classification and localization through spatial and channel reconstruction.

More »

Fig 11 Expand

Table 1.

Training configuration and experimental settings.

More »

Table 1 Expand

Table 2.

Comparison of our RGS-YOLO with the classic methods.

More »

Table 2 Expand

Table 3.

Comparison of ablation experimental results.

More »

Table 3 Expand

Table 4.

Ablation experiments on the public dataset.

More »

Table 4 Expand

Table 5.

Model comparison experiments on the public dataset.

More »

Table 5 Expand

Fig 12.

Detection results comparison.

Each row represents a different input image, and each column corresponds to the detection results produced by a different model.

More »

Fig 12 Expand