Waymo Motion dataset incorrect

[daniel-bogdoll]

The Waymo Motion dataset is listed with camera dn lidar as sensors. The dataset does not provide the sensor information though, it only provides the abstracted 3D bounding boxes of vehicles, pedestrians and cyclists. What would be the best way to integrate such a dataset? Argoverse Motion Forecasting should be the same, as well as INTERACTION etc.

    {
        id: "Waymo Open Motion", //07.08. fertig
        href: "https://waymo.com/open/data/motion/",
        size_hours:  "574",
        size_storage: "",
        frames: "20670800", //calculated
        numberOfScenes: "103354",
        samplingRate: "10",
        lengthOfScenes: "20",
        sensors: "camera, lidar",
        sensorDetail: "5x cameras, 5x lidar, ",
        benchmark: "motion prediction, interaction prediction",
        annotations: "3d bounding boxes, 3d hd map information",
        licensing: "freely available for non-commercial purposes",
        relatedDatasets: "Waymo Open Perception",
        publishDate: "2021.03",
        lastUpdate: "2021.09",
        relatedPaper: "https://arxiv.org/pdf/2104.10133.pdf"
    },

[daniel-bogdoll]

Maybe there should be a label if the raw data is provided or only processed one in the form of 2D/3D bounding boxes. This is different from benchmarks, where bounding boxes etc are provided additionally on the raw data

[daniel-bogdoll]

I think a differentiation between raw and processed data would be a good solution

[Tracy1024]

Ja, raw and processed data is a good solution.