If your Vaion vcore deployment's analytics are detecting objects such as people or vehicles that are not present ("Spurious detections") or failing to detect objects that are present in the video ("Missed detections") then the best thing to do is to submit the problem video to Vaion for further analysis. In most cases we will be able to use these videos to train our analytics to provide more accurate object detection in future. We would also request that you submit a log bundle from your vcore server so that we can check the video streams coming from the camera in question to make sure the camera is providing suitable video for optimum analytic performance.
In case of spurious detections, please export a sample of video from the camera during which spurious/false detections are happening. Ideally there should be as few as possible detectable objects (such as people, vehicles) in this video, and a length of approximately 30 minutes would be best for our training purposes.
Please export the video from your vcore and submit it to us at the following URL;
Submitting a file will give you a code, you can then raise a vcare case and provide this code, together with a brief description of when the detections occurred, and other objects present in the video, and which camera the video came from. Please also upload the logs to the vcare case.
If objects were not detected that should have been, please export the video plus a couple of minutes either side of the missed event, and upload it to the following url;
Submitting a file will give you a code, you can then raise a vcare case and provide this code, together with a brief description of which object was missed and at what time into the exported video the missed object appears, and which camera the video came from. Please also upload the logs to the vcare case.