search

sciai-lab / batchlib

Batch procesing for high-throughput screening data
MIT License
3 stars 0 forks source link

Multiscale images #35

Closed tischi closed 4 years ago

tischi commented 4 years ago

@constantinpape Just to keep track of when the multi-scale images will be generated by the pipeline. Currently they are not yet, right?

constantinpape commented 4 years ago

Just to keep track of when the multi-scale images will be generated by the pipeline. Currently they are not yet, right?

Multi-scale images are generated by the pipeline since e26132369. Most plates processed on the hci server has been started before that was merged, but for all future plates we will have it.

tischi commented 4 years ago

@imagirom Could you run the titration plate with such a pipeline (see above)?

constantinpape commented 4 years ago

Recomputation is done.

tischi commented 4 years ago

@constantinpape This one is still singe scale: data-processed-seg-new/titration_plate_20200403_154849 Maybe other folder?

constantinpape commented 4 years ago

This one is still singe scale: data-processed-seg-new/titration_plate_20200403_154849 Maybe other folder?

please read in slack for details, I posted the new path there

constantinpape commented 4 years ago

Can you load the mult-scale now? Do you think it's working properly?

tischi commented 4 years ago

I can load it, but the shift between the different resolutions somehow got worse... Do you want to quickly zoom?

constantinpape commented 4 years ago

Ok, nvm

constantinpape commented 4 years ago

These are the down-sampled sizes I produce:

/serum                   Group
/serum/s0                Dataset {930, 1024}
/serum/s1                Dataset {465, 512}
/serum/s2                Dataset {232, 256}
/serum/s3                Dataset {116, 128}

This is the down-sampled sizes from BDV:

/t00000/s02/0            Group
/t00000/s02/0/cells      Dataset {1/Inf, 930/Inf, 1024/Inf}
/t00000/s02/1            Group
/t00000/s02/1/cells      Dataset {1/Inf, 465/Inf, 512/Inf}
/t00000/s02/2            Group
/t00000/s02/2/cells      Dataset {1/Inf, 232/Inf, 256/Inf}

So the sizes in x and y are the same. But is there maybe a problem in your code due to the singleton z axis? (Or rather because we don't have the singleton z axis?)

tischi commented 4 years ago

I think I fixed it. It was the issue that I suspected with tiling many of them.

Something else, could do downsample even one more time? If think when zooming out to see the whole plate this may help somewhat.

constantinpape commented 4 years ago

Something else, could do downsample even one more time? If think when zooming out to see the whole plate this may help somewhat.

Sure, no problem.

constantinpape commented 4 years ago

So I think this is fixed now. Closing the issue for now, feel free to reopen if anything else is coming up.