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

Table 1

Results of comparison with other deep learning-based classification baselines for railway image classification.

	Mean precision (%)	Mean recall (%)	Mean F1 measure (%)	FPS	Model size (M)
Proposed	96.88	96.82	96.85	72.18	0.90
ViT	97.51	97.48	97.49	38.94	85.80
MobileNetV2	96.25	96.19	96.22	59.58	2.23
ResNet-50	96.66	96.24	96.64	52.23	23.51