Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Problem counting the vehicles (Wrong direction is counted up) #501

Open
dennbas opened this issue Nov 26, 2021 · 2 comments
Open

Problem counting the vehicles (Wrong direction is counted up) #501

dennbas opened this issue Nov 26, 2021 · 2 comments

Comments

@dennbas
Copy link

dennbas commented Nov 26, 2021

Hello together,

I'm running Opendatacam and Yolov4 on Jetson AGX Xavier with this configuration "config.json" (.json renamed to .txt).
I have currently this problem:
In this video I have two counting lines that have different direction. In some scenes the car was also counted additional in the wrong driving direction. I could not figured out, what kind of error can this be.

demo.mp4
pathfinder.mp4
@vsaw
Copy link
Collaborator

vsaw commented Nov 27, 2021

My guess is, that if you evaluate the tracker data you will find that the center of the bounding box will be moving backwards because of inaccuracies in the detection of the object (notice how the bounding box is jumping even when the car is not moving?)

Currently there is no way to define a "minimum travel distance" or "hysteresis" to prevent this.

However have you tried to set up one bidirectional counting line, at a location where the vehicles are still moving sufficiently fast, instead of two? As it gives you the direction of the vehicle in the JSON/CSV file you should be able to collect the same data but possibly without the errors.

@haviduck
Copy link

My guess is, that if you evaluate the tracker data you will find that the center of the bounding box will be moving backwards because of inaccuracies in the detection of the object (notice how the bounding box is jumping even when the car is not moving?)

Currently there is no way to define a "minimum travel distance" or "hysteresis" to prevent this.

However have you tried to set up one bidirectional counting line, at a location where the vehicles are still moving sufficiently fast, instead of two? As it gives you the direction of the vehicle in the JSON/CSV file you should be able to collect the same data but possibly without the errors.

yeah, ill bet my hat that you right. i faced this when i used the previous coordinate against the current coordinate to check tripline intersection. i store around 20 frames of trajectoryhistory, so my solution was to check first recorded coordinate against the current one instead of using only the previous one. yolo bboxes be jumpy...

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants