search

quantumjot / btrack

Bayesian multi-object tracking
https://btrack.readthedocs.io
MIT License
311 stars 50 forks source link

Division event detected but no children #111

Open nitinsai20 opened 2 years ago

nitinsai20 commented 2 years ago

Hi,

I'm trying to use bTrack on cellpose cytoplasmic segmentation(no nuclei data) and visualize using napari. When I visually check the division events are detected fine (the parent number is assigned to one of the daughter cells). However, when I check for the children of those dividing cells in the tracker data I do not see any children. Can you help me with this, please? I attached the cell_config as a text file. I am happy to provide more information. Thank you.

cell_config -NSB.txt

quantumjot commented 2 years ago

Hi @nitinsai20 - thanks for the report, and sorry for the delay in getting back to you. Can you explain a bit more? Are the divisions detected, it's just that information is missing from the objects?

nitinsai20 commented 2 years ago

Hello,

Thanks for your reply.

Yes, the divisions are detected which I confirmed visually using Napari. But the information is not available in the objects. For example, here cell 39 divides into 39 and 93 cells (image 1 attached). But when I look for the number of children for cell 39 it has none (image 2 attached). Hope this helps and I can provide more information if needed.

Cell_39 Cell_39_dic

quantumjot commented 2 years ago

On a quick inspection it looks like you will need to adjust some parameters. The sigma values for P and R are very high in your config - is there a reason for this? Also, the prob_not_assign is quite low. How does the default config perform? If the divisions are correctly identified, both daughter cells should have new and unique IDs, and the children field will be populated in the parent cell. You should also see that there are Fates.DIVISION hypotheses accepted in the log.