Jitter vs. FTM

When Jitter was first released, I thought that it might be an attractive tool not only for processing of video, but also for advanced data processing, going beyond the list magic made possible by Peter Elsea’s eminent Lobjects.

Yesterday and today I have spent quite a bit of work on this, with limited success. I am currently working on a patch using cv.jit for video tracking. As the tracking data already was prepared as jitter matrixes, it made sense to try to stay in Jitter for further development.

Here are two examples of the kind of problems I have been struggling with:

• Amplification of signals using jit.op @op * @val 4. fails, because jit.op refuse to multiply by more than 1.
• I needed to combine two matrixes, one of type float32 and the other of type char. However it turns out that when casting the char matrix to float32, the index values (1, 2, 3,... that initially were in the range 0-255) gets rescaled to 0-1, rendering it totally useless for what I want to do.

Eventually I just gave up and decided trying to use FTM instead. That turned out to be way easier.

The only downside is that the module will then not make it into the main Jamoma distro in the near future, unless I can convince the rest of the developers that we should permit FTM dependencies.