Training custom dataset using Histogram of Gradients in java [closed]
Hello, anybody know how to create own dataset or own training custom dataset for HOG in java ? i really need that for my research to detect 4 gesture hand using custom dataset for HOG thank you before.
what do you want to do with it ? classification ? detection ?
please explain a bit more.
Actually i need this for hand detection
But i want to create my own dataset not from other people dataset. Can i generate my own .xml or .yaml dataset ?
maybe you're confusing something ?
the HOGDetector can only detect a single class / gesture. (it gives your the rect detected from an image of arbitrary size)
if you want to classify, which of 4 gestures is in the image, you need cropped, equal sized images, and the outcome of the prediction will be: #3
different ideas, right ?
to train the 1st, you probably do not want to reinvent it in java, when there is a readymade train program in c++ for this
I dont get it about needed to cropped, btw in classify with SVM from HOG there are positif and negative sample, for 4 gesture are there will become 4 dataset ? or just 1 ? I need your explanation about this sir thank you before keep replying my post.
Hello berak can you give to me your contact ? of some social media, i have something to ask you in real time.
oh noes, no social media (i abhorr that deeply.)
but come out on irc ? #opencv on freenode (mcclintic there)
What the name of IRC Channels ?
opencv ...
Thanks before berak