Visualisation flow graph
This week I have primarily been working on a way to visualise the output from the FAM cyclic spectrum estimator block.
The following flowgraph examples/amc_famvis.grc is used to initialise the famvis 3D visualisation block I made, making use of 8PSK to test the visualisation. I made use of Qwtplot3D, so it is necessary to install that to use the visualisation.
- https://www.tensorflow.org/versions/r0.10/get_started/os_setup.html#pip-installation – First install TensorFlow 0.10
- Install qwtplot3d (in Arch Linux, you can do: “pacman -S qwtplot3d” to install)
- Install my fork of gr-specest from https://github.com/chrisruk/gr-specest
- git clone my gr-inspector fork from : https://github.com/chrisruk/gr-inspector and switch to the dev_amc branch
- Follow the readme to install
- Currently you can’t zoom or rotate the 3D graph, I’m just trying to fix this issue (the flow graph will terminate if you click on the animated graph)
- I need to enable the graph to take arbitrary size FAM input, so that there will be parameters to take different vector lengths
- Axis need to be labelled based on parameters
Videos of different modulation schemes from the famvis block
Fixing RNG in model generation
I also noticed there was a bug in the random number generation in my model generation code, because I am using multiple python processes to run flow-graphs simultaneously, the processes run at the same time were obtaining the same random number sequence. I have fixed this by reseeding the random number generator for each process. This has altered the accuracy so I’m currently still working on the FAM & CNN model generation code.