A New Video-Based Crash Detection Method: Balancing Speed and Accuracy Using a Feature Fusion Deep Learning Framework

Table 1

Performance of crash appearance feature extractors.

Model name	True positive (TP)	False negative (FN)	False positive (FP)	True negative (TN)	Accuracy (%)
VGG-16	1056	144	273	927	82.63
ResNet-50	1087	113	233	967	85.58
VGG-16 + SE	1075	125	251	949	84.33
VGG-16-CBAM	1103	97	231	969	86.33
ResNet-50 + SE	1132	68	214	986	88.25
ResNet-50- + CBAM	1135	65	171	1029	90.17