I work on a project where user draw a predefined shape, then program scan source image and try to predict this shape on image so program then can use this shape just like an aruco marker.
This is the shape i try to predict and use as marker:
First time i try to predict shape with match shape function but precition fails because of shape from source mostly is distort or perspective is not correct.
Then i try SVM training with 1000 images. This time prediction is more good compare to match shape but not enough accurate.
My question is; I'm on right way? and should continue with SVM training, or there is a better way to predict this shape? Or maybe i should change shape?
Please help me in this regard.
And sorry for my bad english.