Paid job. Developing a Jitter external to specs,

Frans-Jan Wind's icon

Hello all,

I'm in need of an jitter external. This external is already an existing external which was written by David Rockeby as part of his SoftVNS library (v.regions). I've managed to create it myself as an Jitter abstraction (see pasted code). But this version seriously lacks performance. This external should accept a 1 plane char (all calculations are on a single plane grey scale). This plane must be divided in N columns and N rows. Which should be set as an argument. The last thing is a calculation of the amount of white (grey) per cell. All combined cells should be send through the object as one list per frame.

As I said before, speed is of the essence. If you think you're the guy / girl for this job, please let me know your price!

You can reach me @ maxmsp@yourule.nl

Best, Frans-Jan Wind

Max Patch
Copy patch and select New From Clipboard in Max.

Martin Beck's icon

Hi, I'm not looking for a job, just want to propose an idea. Not sure if this is compliant with your speed requirements and other details.
As far as I have understood you are looking for something like the average of values in different subsets of a matrix like the average intensity of adjacent non overlapping receptive fields.
If you could work without the requirement of non overlapping you could do the following:
1) Do blurring of your input (e.g. jit.gl.slab.gauss6x.maxpat, adjust Gauss width according to your number of rows and columns)
2) Resample the blurred result to the required rows and columns by [jit.matrix 1 char columns rows]
3) jit.iter to obtain the list

Pedro Santos's icon

I'd love to earn some money, but Cycling74 has already done it ;-)
jit.dimop + jit.spill is a much easier implementation than your approach and the performance is also better. I don't know if a new external with a different approach could have a much improved performance...

Max Patch
Copy patch and select New From Clipboard in Max.

Frans-Jan Wind's icon

@MARTIN BECK & @PEDRO SANTOS
Thank you for your hand outs. Both methods made me look completely different at this challenge. I feel like getting a head start.

Pedro Santos's icon

For your reference, in the computer I made the previous patch:

Your solution: 33 fps
My solution with @op +: 55 fps
My solution with @op avg: 65 fps