Submitted by
Marin Oršić
Authors
In Defense of Pre-trained ImageNet Architectures for Real-time Semantic Segmentation of Road-driving Images
Venue
CVPR 2019
Abstract
SN_RN152pyrx8_RVC
Runtime
1.00 s
External data used
ImageNet, ADE20k, KITTI, MVD, ScanNet, VIPER, WildDash2
Links
Number of submissions
1

Detailed results
Class IoU (day) IoU (sunset) IoU (rain) IoU (snow) IoU (night) IoU (mean)
Sky95.795.495.395.494.195.2
Road87.694.795.796.796.494.2
Sidewalk79.280.268.487.988.180.8
Terrain63.468.272.571.070.869.2
Tree81.379.580.780.578.980.2
Vegetation71.960.769.067.267.167.2
Building82.482.578.884.782.982.3
Infrastructure44.756.352.259.160.654.6
Fence32.723.723.831.424.227.1
Billboard56.558.356.343.353.853.6
Trafficlight69.565.364.274.261.767.0
Trafficsign47.555.679.566.072.064.1
Mobilebarrier29.67.123.510.89.716.1
Firehydrant45.852.468.259.564.858.1
Chair12.525.95.333.742.624.0
Trash35.219.512.029.430.025.2
Trashcan66.449.747.647.963.154.9
Person83.966.676.972.566.973.3
Motorcycle96.177.83.060.062.359.8
Car95.096.294.698.797.396.4
Van71.54.032.116.427.930.4
Bus76.476.590.452.693.677.9
Truck92.781.785.290.682.686.6
Mean66.059.959.862.164.862.5