Submitted by
Oliver Zendel
Authors
Porzi, Lorenzo and Rota Bulò, Samuel and Colovic, Aleksander and Kontschieder, Peter
Venue
The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2019
Abstract
Seamless Scene Segmentation Resnet101, pretrained on Imagenet; supplied with altered MVD to include WildDash2 classes; does not contain other RVC label policies (i.e. no ADE20K/COCO-specific classes -> rvcsubset and not a proper submission); Trained and Submitted by Oliver Zendel for RVC
Runtime
500 ms
External data used
model pre-trained on ImageNet; regular training on altered MVD to contain WildDash2 labels (van, pickup) (20k frames) no other dataset (i.e. no VIPER frames)
Links
Number of submissions
1

Detailed results
Class IoU (day) IoU (sunset) IoU (rain) IoU (snow) IoU (night) IoU (mean)
Sky93.592.587.287.550.682.3
Road77.483.492.287.583.484.8
Sidewalk66.059.249.156.346.255.3
Terrain47.260.515.329.939.638.5
Tree0.00.00.00.00.00.0
Vegetation22.613.825.520.415.519.6
Building77.271.966.573.356.169.0
Infrastructure0.00.00.00.00.00.0
Fence26.321.114.616.914.118.6
Billboard32.529.131.723.430.429.4
Trafficlight69.165.972.376.458.868.5
Trafficsign23.762.167.827.165.649.2
Mobilebarrier0.10.10.20.00.00.1
Firehydrant52.347.166.272.661.359.9
Chair0.00.00.00.00.00.0
Trash0.00.00.00.00.00.0
Trashcan52.826.727.741.537.337.2
Person74.748.259.172.159.162.6
Motorcycle4.018.129.138.017.721.4
Car53.044.655.953.139.349.2
Van40.30.918.19.90.013.8
Bus16.223.213.63.523.115.9
Truck72.158.257.760.226.554.9
Mean39.235.937.036.931.536.1