Benchmarking using kitti dataset
Description
Purpose
Once the tracker is developed we want to know how it compares to the state of the art trackers. This can be done if we use one of the open source datasets and integrate it with our repository to run our tracker
Update: ARM will contribute their general purpose benchmarking tool as part of this ticket. This can be extended as needed to support tracking evaluation as part of a different ticket.
Desired Behavior
Identify and integrated a open source dataset
Update: This ticket will add support for using KITTI dataset for the purpose of benchmarking
Definition of Done
It should be possible to,
-
Pass point cloud from the beginning of the perception pipeline and produce tracksdetections from it -
Use ground truth detections from the dataset -
Produce some metrics from the above two steps
Edited by Gowtham Ranganathan