Datasets:

harrypotter1501
/

cosmos-trajectory

Tasks:

Time Series Forecasting

Languages:

ArXiv:

License:

Dataset card Files Files and versions Community

harrypotter1501 commited on Apr 22

Commit

eedae63

·

1 Parent(s): 5c78a39

Update readme

Files changed (1) hide show

README.md +1 -1

README.md CHANGED Viewed

@@ -11,7 +11,7 @@ pretty_name: COSMOS Trajectory
 ## Description
-This dataset comprises real-world pedestrian and vehicle trajectories collected from the [COSMOS Testbed](https://www.cosmos-lab.org/). The trajectories are captured at an intersection in New York City using a high-elevation camera, collected using the tracking-by-detection paradigm with a fine-tuned [YOLOv8](https://docs.ultralytics.com/models/yolov8/) object detection model and the [BoT-SORT](https://github.com/NirAharon/BoT-SORT) tracking algorithm.
 Complete trajectories of 176,000 pedestrians and 215,000 vehicles, suitable for statistical analyses. In the context of trajectory forecasting, the dataset offers 510k pedestrian and 250k vehicle trajectories collected over a 30-day period. They are processed into 10k segments of 20-frame (8 seconds) scenes for model training and evaluation.

 ## Description
+This dataset comprises real-world pedestrian and vehicle trajectories collected from the [COSMOS Testbed](https://www.cosmos-lab.org/). The trajectories are captured at an intersection in New York City using a high-elevation camera using the tracking-by-detection paradigm with a fine-tuned [YOLOv8](https://docs.ultralytics.com/models/yolov8/) object detection model and the [BoT-SORT](https://github.com/NirAharon/BoT-SORT) tracking algorithm.
 Complete trajectories of 176,000 pedestrians and 215,000 vehicles, suitable for statistical analyses. In the context of trajectory forecasting, the dataset offers 510k pedestrian and 250k vehicle trajectories collected over a 30-day period. They are processed into 10k segments of 20-frame (8 seconds) scenes for model training and evaluation.