Recordings and Datasets
Sample recordings
Those recordings are available in multiple formats: RAW (EVT2 or EVT3), DAT and HDF5 event files.
Name |
Screenshot |
Description |
Sensor |
Duration |
Recordings |
---|---|---|---|---|---|
spinner |
Turning Spinner |
Gen3.0 |
5 s |
||
hand_spinner |
Turning 3-Blade Hand Spinner |
Gen3.1 |
5 s |
||
laser |
High-speed laser motion |
Gen4.1 |
3 s |
||
80_balls |
Falling 80 balls |
Gen3.0 |
6.3 s |
||
195_falling_particles |
Falling 195 particles of different sizes (synthetic data) |
Gen3.1 |
32 ms |
||
monitoring_40_50hz |
Object vibration at 40 and 50Hz |
Gen3.0 |
6 s |
||
sparklers |
High-speed scatter-like motion of small particles |
Gen3.0 |
95 ms |
||
200_jets_at_200hz |
200 horizontal fluid jets |
Gen3.1 |
2 s |
||
pedestrians |
Pedestrians in multiple directions |
Gen4.1 |
30 s |
||
traffic_monitoring |
Cars moving on a highway (static camera) |
Gen3.0 |
28 s |
||
driving_sample |
Driving on a street road (moving camera) |
Gen4.1 |
12.5 s |
||
box |
Camera moving around a box (to use with metavision_model_3d_tracking sample) |
Gen3.1 |
28.6 s |
||
cube |
Camera moving around a cube (to use with metavision_model_3d_tracking sample) |
Gen3.1 |
17.7 s |
||
marker |
Camera moving around a marker (to use with metavision_model_3d_tracking sample) |
Gen3.1 |
22.3 s |
||
active_marker |
Active Marker moving in front of the camera (to use with Active Marker samples: 2D tracking and 3D tracking) |
IMX636 |
31.6 s |
||
aruco_marker |
ArUco marker moving in front of the camera (to use with metavision_aruco_marker_tracking sample) |
IMX636 |
16.2 s |
Datasets
Dataset |
Application |
Sensor |
Resolution |
Size |
Format |
---|---|---|---|---|---|
Automotive detection |
Gen4.0 |
1280×720 |
1.6TB compressed 3.5TB uncompressed |
Recording in DAT format Labels in numpy format |
|
Corner detection |
Gen3.0 ATIS |
480×360 |
6GB compressed 11GB uncompressed |
Recording in DAT format |
|
Car classification |
Gen1.0 |
crops from 304×240 |
280MB compressed |
Recording in DAT format |
|
Automotive detection |
Gen1.0 |
304×240 |
200GB compressed 750GB uncompressed |
Recording in DAT format Labels in numpy format |
|
Gen4.0 |
1280×720 |
24GB uncompressed |
Recording in DAT format Labels in numpy format |
||
Gesture classification |
Gen3.1 |
640x480 |
113GB uncompressed |
Recording in DAT format Labels in numpy format |
Precomputed Datasets
Those datasets were precompiled in an HDF5 tensor file format for faster training and smaller disk usage.
Dataset |
Preprocessing |
Sensor |
Resolution |
Size |
Format |
---|---|---|---|---|---|
Gen4.0 |
1280×720 down to 640x360 |
211.3GB |
Labels in numpy format |
||
Multi Channel Time Surface |
Gen4.0 |
1280×720 down to 640x360 |
288.4GB |
Labels in numpy format |
|
Gen4.0 |
1280×720 down to 640x360 |
86GB |
Labels in numpy format |
||
Gen1.0 |
304×240 |
59.3GB |
Labels in numpy format |
||
Histogram quantized |
Gen3.1 |
640x480 |
3.2GB |
Labels in numpy format |