[BUG] Cellfinder issues with self-trained model

[Eddymorphling]

Hi folks,
I am using a self trained Cellfinder model and brainmapper CLI to run registration and cell detection. Both registration and cell detections works well but the cell - classification step gives me an error. I get the error mentioned below. Not sure how to solve the issue. Just an FYI my self-trained model gets saved as "model.keras" not sure if this is the issue. Any suggestions? The brainmapper CLI that I am running is below. Thank you

Brainmapper CLI -

brainmapper -s $signal_folder -b $background_folder --trained-model $trained_model -o $output_folder_parent -v 7 2.7 2.7 --orientation sal --soma-spread-factor 1.4 --threshold 8 --soma-diameter 15 --atlas perens_lsfm_mouse_20um --ball-xy-size 2 --ball-z-size 4 --ball-overlap-fraction 0.4 --batch-size 1024 --n-free-cpus 62

Error Message

Traceback (most recent call last):
  File "/home/mvn/conda/envs/BG_test/bin/brainmapper", line 8, in <module>
    sys.exit(main())
             ^^^^^^
  File "/home/mvn/conda/envs/BG_test/lib/python3.12/site-packages/brainglobe_workflows/brainmapper/main.py", line 94, in main
    run_all(args, what_to_run, atlas)
  File "/home/mvn/conda/envs/BG_test/lib/python3.12/site-packages/brainglobe_workflows/brainmapper/main.py", line 168, in run_all
    points = classify.main(
             ^^^^^^^^^^^^^^
  File "/home/mvn/conda/envs/BG_test/lib/python3.12/site-packages/cellfinder/core/classify/classify.py", line 71, in main
    if trained_model and trained_model.suffix == ".h5":
                         ^^^^^^^^^^^^^^^^^^^^
AttributeError: 'str' object has no attribute 'suffix'

[IgorTatarnikov]

Hi @Eddymorphling this was fixed in #471, however this hasn't been released yet.

@alessandrofelder can we release v.1.4.0alpha0?

[alessandrofelder]

Fine by me. Are you happy to do this, @IgorTatarnikov ?

[Eddymorphling]

Thank you!

[Eddymorphling]

Perfect, just downloaded cellfinder 1.4.0. Everything works like a charm! Thanks! @IgorTatarnikov @alessandrofelder

Also this PR to run 2d/3D filtering on Pytorch is much faster and a beautiful idea!