#5 Possible bug: Arena definition frame creates a blind spot when contains the tracked animal
I might have found an issue with arena definition in the last version of ToxTrac (2026 1.4)
We are tracking a single cocroach in an open field arena. I noticed that software consistently lost a track of animal in one specific location in a video where each frame contained a roach. This did not happen in a video starting with several empty frames where the roach was introduced later. Moreover, after several tests I found that the location of blind spot in a problematic video depends on which frame is used to modify the arena.
Steps to reproduce:
1) Define arena using a frame with an animal visible
2) Run tracking and observe that detection consistently fails whenever animal is in the location used in area definition frame.
3) Define arena using a frame with animal in a different location.
4) Blind spot will now move to that location.
I uploaded two screenshots of arena definition window that show different frames and moving location of the blind spot. Change in arena definition also changes visibility statistic (some of the locations may be preferred and if that's the case, visibility drops to 0.2 or below because animal is no longer detected inside the blind spot).
The solution seems to be to have an empty frame inside the video that can be used to define arena. We are also using a temporary fix for already recorded videos to define arena using in-transit locations (sides of open fields instead of corners in our case) because that produces very brief detection failure.
I am happy to provide videos/project files/additional information if that would help to reproduce the issue.
Best,
OLga
Related
Hi, yes I see the problem.
There seem to be an issue with your arena definition stage. You should be able to tune the arena definition settings to include the entire arena. Specifically the brightness threshold values of the arena.
If you are not able to solve the issue, please send me an email. So we can test your video and configuration.
Thank you for getting back to me. The issue is not the arena per se. We are able to include the entire arena and to select the brightness values that detect the roach.
The issue is that unless the frame chosen for identifying the arena is empty, the location where the animal is on that frame creates a blind spot in which the target will be consistently lost. This can be tested by using a different frame—the blind spot will move to a new location.
Does this make sense?
Olga
Hi, yes I see the problem.
There seem to be an issue with your arena definition stage. You should be able to tune the arena definition settings to include the entire arena. Specifically the brightness threshold values of the arena.
If you are not able to solve the issue, please send me an email. So we can test your video and configuration.
[tickets:#5] Possible bug: Arena definition frame creates a blind spot when contains the tracked animal
Status: open
Milestone: 2026 v1
Created: Thu Jul 30, 2026 08:11 PM UTC by Olga F Lazareva
Last Updated: Thu Jul 30, 2026 08:11 PM UTC
Owner: nobody
Attachments:
I might have found an issue with arena definition in the last version of ToxTrac (2026 1.4)
We are tracking a single cocroach in an open field arena. I noticed that software consistently lost a track of animal in one specific location in a video where each frame contained a roach. This did not happen in a video starting with several empty frames where the roach was introduced later. Moreover, after several tests I found that the location of blind spot in a problematic video depends on which frame is used to modify the arena.
Steps to reproduce:
1) Define arena using a frame with an animal visible
2) Run tracking and observe that detection consistently fails whenever animal is in the location used in area definition frame.
3) Define arena using a frame with animal in a different location.
4) Blind spot will now move to that location.
I uploaded two screenshots of arena definition window that show different frames and moving location of the blind spot. Change in arena definition also changes visibility statistic (some of the locations may be preferred and if that's the case, visibility drops to 0.2 or below because animal is no longer detected inside the blind spot).
The solution seems to be to have an empty frame inside the video that can be used to define arena. We are also using a temporary fix for already recorded videos to define arena using in-transit locations (sides of open fields instead of corners in our case) because that produces very brief detection failure.
I am happy to provide videos/project files/additional information if that would help to reproduce the issue.
Best,
OLga
Sent from sourceforge.net because you indicated interest in https://sourceforge.net/p/toxtrac/tickets/5/
To unsubscribe from further messages, please visit https://sourceforge.net/auth/subscriptions/
Related
Tickets: #5