Datasets

RoadObstacle21

Obstacle track: obstacle segmentation with the road as region of interest

• 412 test images with pixel-level annotations of resolution 1920x1080
• 30 extra images published with pixel-level annotations
• the obstacles in this dataset can be understood as anomaly objects as well
• object types: e.g. stuffed toys, sleighs, tree stumps, ...
• obstacles appear at different distances (one distance per image)
• different road surfaces, lighting and weather conditions available

RoadAnomaly21

Anomaly track: general anomaly segmentation in full street scenes

• 100 test images with pixel-level annotations of resolution 2048x1024 and 1280x720
• 10 extra images published with pixel-level annotations
• many anomalous objects per image, e.g. herd of sheeps
• anomalies can appear anywhere in the image
• anomalies widely differ in size (from 0.5% to 40% of the image)
• images were collected from web resources and therefore depict a wide variety of environments

The Fishyscapes Lost and Found dataset is an anomaly detection for semantic segmentation dataset based off of the Cityscapes dataset. The button below will direct you to the Lost and Found website, where the dataset is available for download. We expect submission of the full Lost and Found dataset to the benchmark to keep exact image names private.

Fishyscapes Lost & Found

• 275 test images from different locations.
• Excludes sequences with children or bikes, as these are not considered anomalies.
• Removes images without objects to focus on relevant data.
• Features small anomalous objects that are not uniformly distributed in the images.
• Enables testing with real images to prevent overfitting on synthetic image processing.