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.

Channel shuffle flowchart.

(a) shows the original grouped convolution, (b) shows different channels being shuffled to different positions after grouped convolution, and (c) shows the result after channel shuffling is completed.

More »

Fig 1 Expand

Fig 2.

SH-encoder module, DW stands for depthwise separable convolution.

More »

Fig 2 Expand

Fig 3.

UPC-SimAM module, DW stands for depthwise separable convolution.

More »

Fig 3 Expand

Fig 4.

SimAM attention.

(a) Channel-wise attention, (b) Spatial-wise attention, (c) Full 3-D weights for attention.

More »

Fig 4 Expand

Fig 5.

SH-DETR architecture.

More »

Fig 5 Expand

Table 1.

Experimental hardware and software configuration.

More »

Table 1 Expand

Fig 6.

Examples of each category in NEU-DET.

More »

Fig 6 Expand

Fig 7.

Distribution of categories in the NEU-DET dataset.

More »

Fig 7 Expand

Fig 8.

Examples of each category in GC10-DET.

More »

Fig 8 Expand

Fig 9.

Distribution of categories in the GC10-DET dataset.

More »

Fig 9 Expand

Table 2.

The detection performance of different categories on NEU-DET dataset.

More »

Table 2 Expand

Fig 10.

Loss function curve and detection precision experimented on NEU-DET dataset.

More »

Fig 10 Expand

Fig 11.

Confusion matrix of the detection results on NEU-DET dataset.

More »

Fig 11 Expand

Fig 12.

Loss function curve and detection precision experimented on GC10-DE dataset.

More »

Fig 12 Expand

Table 3.

The detection performance for different dataset.

More »

Table 3 Expand

Table 4.

The detection performance for different categories on GC10-DE dataset.

More »

Table 4 Expand

Table 5.

The detection performance for different categories on welding defect dataset.

More »

Table 5 Expand

Fig 13.

Examples of images processed with different data augmentation methods.

More »

Fig 13 Expand

Fig 14.

Loss curves and detection precision for processed images on the GC10-DE dataset.

More »

Fig 14 Expand

Fig 15.

Histogram of model accuracy training results.

More »

Fig 15 Expand

Table 6.

Comparison of our proposed model with other state-of-the-art models on NEU-DET dataset.

More »

Table 6 Expand

Table 7.

Comparison of detection methods by defect category on NEU-DET dataset.

More »

Table 7 Expand

Table 8.

Comparison of detection methods by defect category on GC10-DET dataset.

More »

Table 8 Expand

Table 9.

Impact of each module on the model in ablation studies.

More »

Table 9 Expand

Fig 16.

Visualization results of different modules in ablation studies.

More »

Fig 16 Expand