Fig 1.
The overall architecture of the network.
Fig 2.
Multi-scale pyramid convolution kernel.
Fig 3.
Grouped convolution.
Table 1.
PyconvResNet50 convolution kernel size information.
Fig 4.
Multi-attention module.
Fig 5.
Channel attention module.
Fig 6.
Spatial attention module.
Table 2.
Details of the three datasets.
Table 3.
Top1 accuracy (%) of convolution combinations with different sizes on CUB-200-2011.
Table 4.
Experimental results of ablation.
Top1 accuracy (%) on CUB-200-2011.
Fig 7.
Back propagation saliency map.
Fig 8.
Visualization of attention.
Fig 9.
The test accuracy curve and train loss curve on CUB-200-2011, Stanford Cars and Stanford Dogs datasets.
Table 5.
Comparison results on CUB-200-2011.
Table 6.
Comparison results on Stanford Dogs and Stanford Cars.