Table 3.
A summary of the available datasets containing radar data and/without other sensors data.
| Dataset | Sensing Modalities | Size | Scenes | Labels | Frame Rate | Radar Signal Representation | Type of Objects | Recording Condition |
Recording Location | Published Year | Availability/LINK |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Nuscenes [41] | Visual Cameras (6), 3D Lidar, and Radars (5) | 1.4 frames (Cameras and Radars) and 390 K frames of Lidar | 1 K | 2D/3D bounding boxes (1.4M) | 1 HZ/10 HZ | 2D radar point clouds | 23 object classes | Nightime/rain and light weather | Boston, Singapore. | 2019 | Public, (https://www.nuscenes.org/download (accessed on 20 January 2021)) |
| Astyx HiRes [43] | Radar, visual cameras, and 3D Lidar | 500 frames | n.a | 3D bounding boxes | n.a. | 3D Radar point clouds | Car, bus, motorcycle, person, trailer, and truck | Daylight | n.a. | 2019 | Public, (http://www.astyx.netaccessed on 30 December 2020)) |
| CARRADA [44] | Radar and Visual camera | 12,726 total number of frames | n.a | Sparse point, bounding boxes, and dense masks. | n.a. | Range-angle and range-Doppler raw radar data | cars, pedestrians, and cyclist | n.a. | Canada | 2020 | To be released |
| [52] | Radar | n.a. | Parking lot, campus road, city road and, freeway | Metadata (Object location and class) | 30 fps | Raw radar I-Q samples | Pedestrian, cyclist, and cars | Under challenging light conditions | n.a. | 2019 | To be released via (https://github.com/yizhou-wang/UWCR (accessed on 5 December 2020) |
| CRUW [98] | Stereo cameras and 77GHz FMCW Radars (2) | More than 400 K frames | Campus road, city street, highway, parking lot, etc | Annotations (Object location and class) | 30 fps | Range-Azimuth maps (RAMaps) | About 260 K objects | Different autonomous driving scenarios, including dark, strong light, and blur | n.a | 2020 | Self-collected |
| [156] | Cameras(2), Lidars(2), radar, gated NIR, and FIR | 1.4 M frames | 10,000 km of driving | 2D/3D (each 100k) | 10 Hz | Radar signal projections | n.a | Clear, nighttime, dense fog, light fog, rain, and snow | Northern Europe (Germany, Sweden, Denmark, and Finland) | February and December 2019 | On request |
| Oxford Robot-Car [170] | Visual cameras (6), 2D & 3D Lidars (4), GNSS, Radar, and Inertial sensors | 240 k (Radar), 2.4 frames (Lidar) and 11,070,651 frames (Stereo camera | n.a. | No | n.a. | Range-Azimuth maps | Vehicles and pedestrians | Direct sunlight, heavy rain, night, and snow | Oxford | 2017, 2020 | Public, (https://ori.ox.ac.uk/oxford-radar-robotcar-dataset (accessed on 7 January 2021)) |
| SCORP [171] | Radar and Visual camera | 3913 frames | 11 Driving sequences | Bounding boxes. | n.a. | SCA, RDA, and DoA tensors | n.a. | n.a. | n.a. | 2020 | Public, (https://rebrand.ly/SCORP (accessed on 27 November 2020)) |