Overview

A Synthetic Dataset with Multiple Uncertainties for Autonomous Driving

MUAD

We introduce MUAD, a synthetic dataset for autonomous driving with multiple uncertainty types and tasks. It contains 10413 in total: 3420 images in the train set, 492 in the validation set and 6501 in the test set. The test set is divided as follows: 551 in the normal set, 102 in the normal set no shadow, 1668 in the OOD set, 605 in the low adversity set and 602 images in the high adversity set 1552 in the low adversity with OOD set and 1421 images in the high adversity with OOD set. All of these sets cover day and night conditions, with 2/3 being day images and 1/3 night images. Test datasets address diverse weather conditions (rain, snow, and fog with two different intensity levels) and multiple OOD objects.

MUAD has seven test sets:

  • normal set: images without OOD objects nor adverse conditions
  • low adversity with OOD set: images containing both OOD objects and medium intensity adversity conditions (either fog, rain or snow)
  • high adversity with OOD set: images containing both OOD objects and high intensity adversity conditions (either fog, rain or snow)
  • normal set no shadow: images without OOD objects nor adversity conditions and we simulate the sun like if it was 12:00 am so the shadow is minimum
  • OOD set: images containing OOD objects and without adversity conditions
  • low adversity set: images containing medium intensity adversity conditions (either fog, rain or snow)
  • high adversity set: images containing high intensity adversity conditions (either fog, rain or snow)

MUAD supports four tasks:

  • Semantic segmentation
  • Depth estimation
  • Object detection
  • Instance detection

Overview of annotated classes:

There are a total of 155 fine grained classes, which are also aggregated in order to facilitate the use along with other datasets, e.g. Citiscapes:

Cityscapes classesMUAD classesnb. of images with the annotations
RoadBots, Tram Tracks, Crosswalk, Parking Area, Garbage - Road, Road Lines, Sewer Longitudinal Crack, Transversal Crack, Road, Asphalt Hole, Polished Aggregate, Vegetation - Road, Sewer - Road, Construction Concrete9055
SidewalkLane Bike, Kerb Stone, Sidewalk, Kerb Rising Edge8948
BuildingHouse, Construction Scaffold, Building, Air Conditioning, Construction Container, TV Antenna, Terrace, Water Tank, Pergola Garden, Stairs, Dog House, Sunshades, Railings, Construction Stock, Marquees, Hangar Airport9089
WallWall1101
FenceConstruction Fence, Fences8622
PoleTraffic Signs Poles or Structure, Traffic Lights Poles, Street lights, Lamp8984
Traffic lightTraffic Lights Head, Traffic Cameras, Traffic Lights Bulb (red, yellow, green)8222
Traffic signTraffic Signs2672
VegetationVegetation9072
TerrainTerrain, Tree Pit8377
SkySky8591
PersonWalker, All colors of Construction Helmet, All colors of Safety Vest, Umbrella, People8843
RiderCyclist, Biker3470
CarCar, Beacon Light, Van, Ego Car9026
TruckTruck5533
BusBus0
TrainTrain, Subway2240
MotorcycleMotorcycle, Segway, Scooter Child2615
BicycleBicycle, Kickbike, Tricycle2816
AnimalsCow, Bear, Deer, Moose603
Objects anomaliesStand Food, Trash Can, Garbage bag352
BackgroundOthers-


Examples


Move the mouse over the semantic segmentation label map, and the corresponding RGB image will appear.


Resources



Paper

Github repository

Challenge

Poster




Download

Terms of use


Copyright for MUAD Dataset is owned by Université Paris-Saclay (SATIE Laboratory, Gif-sur-Yvette, FR) and ENSTA Paris (U2IS Laboratory, Palaiseau, FR).

If you need MUAD Dataset, please Click and Fill in this Google form. We provide you with permanent download links as soon as you finish submitting the form.




Acknowledgments


We gratefully acknowledge the support of DATAIA Paris-Saclay institute which supported the creation of the dataset (ANR–17–CONV–0003/RD42). We are also grateful to Yan Chen for his help with the early processing of the dataset, as well as the many staff who worked hard to generate the dataset.

Citation

If you use MUAD in your work, please cite this publication:

Contributors

Gianni Franchi

U2IS, ENSTA Paris, Institut Polytechnique de Paris

Xuanlong Yu

SATIE, Paris Saclay University
U2IS, ENSTA Paris, Institut Polytechnique de Paris

Andrei Bursuc

valeo.ai

Ángel Tena

Anyverse

Rémi Kazmierczak

U2IS, ENSTA Paris, Institut Polytechnique de Paris

Séverine Dubuisson

LIS, Aix Marseille University

Emanuel Aldea

SATIE, Paris Saclay University

David Filliat

U2IS, ENSTA Paris, Institut Polytechnique de Paris

contact

If you have any questions (about the dataset or this website), please feel free to contact (replace at by @):

Gianni Franchi

gianni.franchi at ensta-paris.fr

Emanuel Aldea

emanuel.aldea at universite-paris-saclay.fr

Xuanlong Yu

xuanlong.yu at universite-paris-saclay.fr