Weird cv.jit.blobs.sort matrix output processing
Hello,
I imagine cv.jit.blobs.sort is a special object, but to refuse to be processed like a normal matrix ? I just want to make a float32 matrix out of it after processing (e.g. [1., 2.] instead of [1, 2].
Can someone look at the outputs of the jit.prints and explain to me what is happening!
Later edit:
I see a [jit.matrix 1 float32 3 1] gives me 0.004, 0.008, 0.012 for 1, 2, 3 which is decent because I can multiply it by 256 (related to bytesize somehow?) and I get the floats I want, but it's still weird
Hi,
It's not a special object; it outputs regular matrices. When converting from char to float matrices, Jitter divides the values by 255, so that the output is contained between 0 and 1. In many situations, this makes sense. For instance, when you are working with images. When you are working with data, like you are now, it makes less sense. After conversion, you can multiply back by 255, or you can convert to long before converting to float. The new object cv.jit.changetype should be useful in a situation like this. Just connect [cv.jit.blobs.sort] -> [cv.jit.changetype long] -> [cv.jit.changetype float32].