Ask Your Question
0

car detection using open cv

asked 2018-04-26 06:42:38 -0600

holger gravatar image

Hello, After reading ,reading and reading and some working code i want to take some steps back and ask some advice. To solve car detection i currently see two way

  1. Training (several) cascade classifiers
  2. Import a dnn with ssd capability and get the detection results from it

Is training classifiers any good for this kind of object(car)? Should i stick to the dnn - this gives me the most accurate results right now.

Thanks a lot, Greetings, Holger

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted
1

answered 2018-04-27 06:22:47 -0600

Simply said

  • Basic cascades, like the ones in OpenCV will not be enough for robust car detection.
  • However there are more advanced cascade based techniques, like ACF / ICF that will do just fine.

Advantage of using DNN, which can be either YOLO or SSD, is that both techniques contain several off-the-shelf models that contain the car class, and thus we can use the models without retraining anything.

edit flag offensive delete link more

Comments

However there are more advanced cascade based techniques, like ACF / ICF that will do just fine Just to clarify, these techniques or not implemented in open cv (yet)?

About DNN: I am using a pretrained model(darknet) - but that has many more classes it will recognize. So i was thinking about retraining(the other question you anwered) that model. Is that still a good idea? Thx, Holger

holger gravatar imageholger ( 2018-04-27 08:12:46 -0600 )edit

Yes and yes.

StevenPuttemans gravatar imageStevenPuttemans ( 2018-04-27 13:54:45 -0600 )edit
1

Thank you. +

holger gravatar imageholger ( 2018-04-28 02:31:02 -0600 )edit

Question Tools

1 follower

Stats

Asked: 2018-04-26 06:42:38 -0600

Seen: 383 times

Last updated: Apr 27 '18