Example: Scikit-image and trackpy (bubble tracking in foams)

[Emmanuelle Gouillart]

Great, thanks! It seems that most processing pipelines rely on first
detecting objects of interest (eg with a segmentation step) and then
tracking such objects. Do you know of any (generic-enough) approach that
would perform both steps at the same time, ie if you know that the same
objects must be found in several images this information can be used for
performing the segmentation?

On Thu, Nov 20, 2014 at 04:05:37PM -0500, François Boulogne wrote:
> Hi Emmanuelle,

> Le 20/11/2014 14:22, Emmanuelle Gouillart a écrit :
> > Very nice! I just tweeted about it :-). 

> :)
> > How good is that trackpy library? On Tuesday I was giving a presentation
> > about scikit-image and someone in the audience asked me if I knew some
> > good tools for 2D/3D + time image processing, for example for tracking
> > particles/cells. I didn't know any, and actually I don't know whether
> > there exist some "classical" and robust algorithms for particle tracking
> > (like median filter for denoising, otsu thresholding, etc.), or if
> > everything is very application-dependent. What is your experience about
> > this?

> I have a great experience and I warmly recommend the library. It's very
> easy to use, I'd say an investment of few hours to go through the
> tutorials and play a bit.

> The documentation is clear and the two mains steps (feature detection
> and feature tracking) are well separated. Trackpy provides an algorithm
> to detect particles, but you can also detect bubbles and probably cells
> if you write your own detection algorithm.
> Also, Trackpy smartly uses Pandas. Thus, it's very easy to manipulate
> the data and store everything in a h5 container.

> I had the occasion to meet Dan. The authors are very responsive and they
> are also on this ML afaik.

> I'm currently using trackpy for another project too.

> Best,