aKzenT's profile - activity

I had the same question and I learned that as Steven says, it depends on your application. The negative images should be representative of your application, so if you want to detect faces in holiday pictures, then having holiday pictures (without faces) should be a good idea. If you are using kinect, maybe indoor pictures might be more representative. Also as I understand it traincascade will use your negative images to subsample the actual negative pictures. However if I understand the source code correct, it will start subsampling with a window size that is equal to the window size of your positive images, so it might be a disadvantage if your negative images have a higher resolution as it will start subsampling with very small regions of your overall image.

I am trying to build a cascade classifier for license plate detection. So far I managed to get some good results, but could be better.

Looking at all license plates that should be recognized, there are some similarities (all are rectangular and with some letters on them), but other than that they differ between countries and sometimes there are various "categories" inside the same country even.

My question is: Should I train for every country and type of license plate separately and then trying to combine the classifiers later or am I likely to get better/same results when training one classifier with the complete set of license plates?

Thanks!