Foreign Object Detection in Railway Images Based on an Efficient Two-Stage Convolutional Neural Network

<div>Schematic visualization of the improved inverted residual unit. Compared with the original inverted residual unit, we use SK convolution (purple block) to replace the depthwise convolution. To further improve the performance of the unit, we employ the CBAM mechanism (green box).</div>

Computational Intelligence and Neuroscience

fig6

Figure 6

Figure 6: Foreign Object Detection in Railway Images Based on an Efficient Two-Stage Convolutional Neural Network 