01-07-2014 12:56 AM
Hello,
you need to check the block diagram - you can see that there is a diagram disable structure, where the image already in memory is used to calculate the Hog descriptors. You need to disable this and enable the "load image" part!
As for tweaking other parameters (i.e. gamma correction, detection windows size, etc...) you can easily implement them in the source code and rebuild the .dll. Tweaking the parameters should give you better results for your application.
Best regards,
K
01-11-2014 04:22 PM
Hello,
Thanks a lot for your help,but can your VI be improved to detect a "rotated" human? ..For example a human that is lying horizontally on the ground.
01-13-2014 01:01 AM - edited 01-13-2014 01:01 AM
Hello,
you can use OpenCV to train your custom HOG descriptors. For example just one of the tutorials:
https://github.com/DaHoC/trainHOG/wiki/trainHOG-Tutorial
Best regards,
K
01-06-2015 12:40 AM
According to the paper published by Dalal and Triggs, they suggest gamma correction is not necessary. I have the doubt about whether correcting gamma is a good option to go for or not. If Gamma correction is necessary what is the gamma value I have to take for better performance. Awaiting for reply.!!! thanks in advance!!