User defined modelling of the temporal behaviour

[dkazanc]

User defined modelling of the temporal behaviour

[paskino]

I guess we should also think whether we want to do this in C or in Python.
If in Python it shouldn't be too difficult.

[dkazanc]

yes, worth talking about it. Probably we should come up with a convenient parametrisation. If in python then this can be similar to this script, where just linear vertical shift is implemented. However the vector of shifts (translations, rotations) can be of any trajectory.

[paskino]

If we want to support a C API (or wrappers for other languages as MATLAB) would be the ability to set each parameter as:

1. float or constant in time
2. a tuple (p0,p1), implying linear change within the time frame
3. a "list" of values the object assumes. In this case there is scope for interpolation when each parameter may vary with a different number of timeframes.

Worth having a look at what the CCPPETMR numerical simulation implementation.
https://youtu.be/UXrAVnTw7lo?t=3752

I suggest to change the current implementation which assumes equal steps to use generic arrays where the value of the parameter at each step is defined.

TODO, update these lines:

1. TomoPhantom/Core/TomoP2DModelSino_core.c

   Line 436 in 8e9d0c3

   for(tt=0; tt < steps; tt++) {

2. TomoPhantom/Core/TomoP2DModel_core.c

   Line 283 in 8e9d0c3

   float distance = sqrtf(pow((x1 - x0),2) + pow((y1 - y0),2));

3. TomoPhantom/Core/TomoP2DModelSino_core.c

   Line 424 in 8e9d0c3

   float distance = sqrtf(pow((x1 - x0),2) + pow((y1 - y0),2));

4. TomoPhantom/Core/TomoP3DModel_core.c

   Line 367 in 8e9d0c3

   float distance = sqrtf(pow((x1 - x0), 2) + pow((y1 - y0), 2) + pow((z1 - z0), 2));