Well I am doing a project in which I want to recognize a object using Kinect and open cv!! But that object have a lot of different sizes!!I could take a lot of photos of that object!! my question is What type of machine learning do you recommend to use in this case and what type of features detection like SURF or others?
If you use SIFT descriptor then you will get a descriptor matrix of 1 x 64 and similarly if you use SURF descriptor then you will get a descriptor matrix of 1 x 128. So, for each image you will get a matrix either of size 1 x 64 or 1 x 128. Then you use this single row matrices as input vectors for SVM training.