Table 1

Pedestrian crossing intention prediction accuracy for different methods on three public datasets.

MethodPIEJAAD allJAAD beh
ACCAUCF1PredskιnRecallACCAUCF1PrecisionRecallACCAUCF1PrecisionRecall
SFGRU ([23])0.820.790.690.670.70.840.840.650.540.840.510.450.630.610.64
I3D ([7])0.810.830.720.600.90.840.80.630.550.730.620.510.750.650.88
TrouSPI-Net ([9])0.880.870.800.770.840.820.770.580.490.700.640.550.760.650.91
MultiRNN ([4])0.830.80.710.690.730.790.790.580.450.790.610.50.740.640.86
D. Yang. et al. ([40])-----0.830.820.630.510.810.620.540.740.650.85
PCPA ([13])0.870.860.77--0.850.860.68--0.580.50.71--
IntFormer ([16])0.890.920.81--0.860.780.62--0.590.540.69--
Yu Yao. et al. ([41])0.840.900.880.96-0.870.700.920.66------
BiPed ([25])0.910.900.850.82-0.830.790.600.52------
Ours0.910.890.840.840.850.890.780.660.720.610.680.630.760.710.81

or Create an Account

Close subscription notice
Close access options