Openv how to recognize a truck or car in a video?

Question

C++ / C#

Good afternoon!
There are several questions on the topic:
1) What algorithm can be used? (Haar could come up, but not the fact that it will recognize a dynamic image, 5k positive and 5k negative ones will be needed for training)
2) It turns out that it is necessary to apply the algorithm twice for each object? (in my head it looks like this: two cascades are set and alternately pass through the image (for a long time, but most likely it will work))
I would be very grateful for any help (pieces of code, a full-fledged source code of the program, advice, a link to a similar question).

Answer 1

Haar passes quickly enough. Most of the windows are discarded at the initial stages of the cascaded filter.
As for recognition, depending on what the initial data will be, haar in most situations does not recognize the car. Let's say if it is rotated and the selection was made for cars from the side, then the haar will almost never recognize it.
As for the codes, I used ready-made binaries from OpenCV.
In build/[your version of the system]/[your version of VS]/bin there
are files:
opencv_createsamples
opencv_haartraining
opencv_traincascade
With the help of these programs, you can train the classifier according to your sample for further use. More information about initial parametersdocs.opencv.org/doc/user_guide/ug_traincascade.html