Railway Subgrade Defect Automatic Recognition Method Based on Improved Faster R-CNN

Table 1

Data distribution in dataset.

Dataset	Mud pumping	Subgrade settlement	Water abnormality	Ballast fouling	Normal subgrade	Total
Training set	360	280	280	320	400	1640
Validation set	45	35	35	40	50	205
Test set	45	35	35	40	50	205
Total	450	350	350	400	500	2050