You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When testing the code file on other videos yesterday (January 18th, 2024) I realized I wrote the temporal redundancy algorithm to handle only one plate area detection (per car) and only one string detection per plate area.
This causes problems because the same voted plate will display on all plate area detections, and the temporal redundancy algorithm will combine all of the different plate detections no matter what region they come from and create a completely wrong plate based on two different data points that it views as one.
Possible fix
This could be fixed by using similar code to what I have written for logging the plates for specific vehicles by using the "track" function from Ultralytics and using their ID number to differentiate detections. I would also split the detections under each vehicle into multiple tracking IDs, and log them separately under each vehicle.
After thinking over the amount of complicated work I’d have to go through to allow for all of this data, and writing complicated code to allow for tracking multiple detections I decided not to do this.
For now unless there are many times where I have multiple unsuccessful detections in videos, then I will just be using my current temporal redundancy algorithm that takes all string detections from all plate area detections and displaying that as the voted plate for each vehicle.
However, I won’t make changes to the code displaying the detection information in case errors come up later down the road, it will be easier for me to identify and fix.
The text was updated successfully, but these errors were encountered:
Issue explanation
When testing the code file on other videos yesterday (January 18th, 2024) I realized I wrote the temporal redundancy algorithm to handle only one plate area detection (per car) and only one string detection per plate area.
Possible fix
This could be fixed by using similar code to what I have written for logging the plates for specific vehicles by using the "track" function from Ultralytics and using their ID number to differentiate detections. I would also split the detections under each vehicle into multiple tracking IDs, and log them separately under each vehicle.
The logs would look something like this:
Roadblocks
After thinking over the amount of complicated work I’d have to go through to allow for all of this data, and writing complicated code to allow for tracking multiple detections I decided not to do this.
For now unless there are many times where I have multiple unsuccessful detections in videos, then I will just be using my current temporal redundancy algorithm that takes all string detections from all plate area detections and displaying that as the voted plate for each vehicle.
However, I won’t make changes to the code displaying the detection information in case errors come up later down the road, it will be easier for me to identify and fix.
The text was updated successfully, but these errors were encountered: