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

Table 3

Model 6 versus the C3D model in detection accuracy and speed.

Model	Train accuracy (%)	Test accuracy (%)	Parameters (MB)	FLOPs/per video (B)	FPS
C3D [30]	99.89	67.22	249.99	574.36	14
Model 6	96.58	87.78	24.22	265.26	30