search

DeepTrackAI / DeepTrack2

DeepTrack2
MIT License
157 stars 47 forks source link

Particle tracing in 3D by holographic data using LODESTAR #213

Open avi-xd opened 1 month ago

avi-xd commented 1 month ago

Hi, I have been using the LODESTAR model to get the traces of particles for my off-axis Holographic data. I used the model and got results quite accurately on the X-Y plane. However, on the Z axis, the model could not predict very efficiently, as it gave sudden jumps (spikes) between the traces (I am uploading the output).

63134570-9623-4df1-b0d9-e86bd4a80680

But then I checked with the example uploaded by you guys in 07.track_3D_holography.ipynb

I got the same results as in the paper https://www.nature.com/articles/s41467-022-35004-y. But when I plotted the traces in 3D, they also had similar jumps (pics attached below).

In 2D (x-y plane), result is

65806f64-c9a9-493b-ad83-26d3c81b87aa

and similar result in 3D looks like

cdd58616-c7f7-4fb0-9252-6468726e9b30

I am not able to figure out what causes these sudden spikes in the results and how to tackle it and get better results. Any help or key insight will be highly appreciated.

I can provide the exact code I have used if needed. Thanks

avi-xd commented 2 weeks ago

@BenjaminMidtvedt any inputs here? Thanks