Get the detection results in each frame and save in database (Solved)


Can I ask if I get the labels and crop the specific areas of detected objects in each frame and save these detection results data in a database (e.g. Minio)




My two cents.

In the sample nvgstiva-app (presuming that is what you are referring too) we kept it simple and the app writes a kitti output (file for every frame with detections there-in) and a all_bbox_00.txt style file if the bbox-file app is set to 1; it will be output in the CWD.

We utilize the kitti output files, parse out the necessary lines and write them as metrics to influxdb. A bit tedious as a file per frames ends up yields 100s of thousands of files, but if done in parallel, it should be fairly easy to parse and send over to a db. But, the all_bbox may work for you as well, depending the type of metrics required.

Hope it helps, cheers.

Thank you very much.