Run Bigstream2 in Pipeline

[anaste123]

Bug report

Hi, I was trying to do registration of our large image data using Bigsteam2 in EASI-FISH Pipeline, I changed most parameters exposed, including:

    --bigstream_global_steps
    --bigstream_local_steps
    --global_iterations
    --local_iterations
    --local_ransac_cc_radius
    --local_ransac_nspots
    --local_ransac_match_threshold
    --local_ransac_diagonal_constraint
    --aff_scale
    --def_scale
    --global_ransac_fix_spot_winsorize_limits
    --global_ransac_mov_spot_winsorize_limits
    --local_ransac_fix_spot_winsorize_limits
    --local_ransac_mov_spot_winsorize_limits

I also tried the parameters you recommended before:

The same image can get good registration results using Bigstream2 Python script, but can not get good results using Bigstream2 in EASI-FISH Pipeline. From my experience, I think shrink_factors, smooth_sigmas from affine_kwargs/deform_kwargs have great influence on the registration results.

Could you please expose those parameters in Registration step? Is there any other parameters exposed that I can use to improve the registration results? Thanks in advance!

Description of the problem

Log file(s)

Environment

• EASI-FISH Pipeline version: [dev]
• Nextflow version: [23.04.1]
• Container runtime: [Singularity]
• Platform: [Slurm]
• Operating system: [CentOS 7.6]

Additional context

[cgoina]

@anaste123 I finally got to this - could you please tell me which parameters you need to be exposed. If used as a library bigstream is much more flexible and we cannot expose every possible parameter, but we want to expose all parameters that are typically used and lead to better results.