Some question about how-to-estimate-vehicle-speed-with-computer-vision.ipynb

[zzt941006]

Search before asking

• I have searched the Roboflow Notebooks issues and found no similar bug report.

Notebook name

notebooks/notebooks
/how-to-estimate-vehicle-speed-with-computer-vision.ipynb

Bug

Hi, I have some problems about estimate vehcle-speed.
First of all:
If I want to test by my video,do you have any experience in setting the values ​​of SOURCE and TARGET according to real w and h?

After that:
how can I show detect classes on frame? I try to modify code: labels.append(f"#{tracker_id} {int(speed)} km/h")
but failed.

Finally I saw some frame car speed greater than 130KM/H, what parameter can I fix to let vehicle speed is more accurate.

Thanks!

Environment

Local
windows 10
py3.8

Minimal Reproducible Example

No response

Additional

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

[SkalskiP]

Hi @zzt941006! Just wondering if you've had a chance to watch my tutorial on YouTube?

[zzt941006]

Hi @zzt941006! Just wondering if you've had a chance to watch my tutorial on YouTube?

Yes I saw this tutorial,
After that I add some code to show classes on every frame.
But I still can't understand how to set TARGET_WIDTH and TARGET_HEIGHT efficiently.
In my case, cars and buses, truncks speed not more than 60km/h， so if use original params such as TARGET_WIDTH = 25
TARGET_HEIGHT = 250 to estimate speed, some cars speed are 100+ km/h, and then, I modify them to 15, 100 seems work.
However, in different scenarios, TARGET_WIDTH and TARGET_HEIGHT should be adaptive and cannot be adjusted manually all the time. What suggestions do you have in this regard?

[SkalskiP]

TARGET_WIDTH and TARGET_HEIGHT must be chosen individually based on the actual dimensions of the road. I read my 25x250 dimensions from Google Maps. The solution proposed in this tutorial is not adaptive and requires manual input.

[zzt941006]

TARGET_WIDTH and TARGET_HEIGHT must be chosen individually based on the actual dimensions of the road. I read my 25x250 dimensions from Google Maps. The solution proposed in this tutorial is not adaptive and requires manual input.

Thanks for apply! Beside this, I also want to know how to evaluate the accuracy of estimated speed.In your tutorial, although the speed is estimated, we cannot judge whether the estimated speed is correct. I find some open dataset, only labeled car's bounding box, rather than speed. So for this problem, do you have any suggestions for me? Looking forward to your reply~