feature: save labeled segmentations rather than binary segmentations #193
Conversation
hollandjg
requested review from
cpaniaguam,
danielmwatkins,
mirestrepo,
mkkim400 and
tdivoll
test/test-IFTPipeline.jl-cli.sh
Outdated
There was a problem hiding this comment.
Are you intending to use a Makefile or something to orchestrate the initialization? What's the plan?
There was a problem hiding this comment.
Nope, this is just a simple test script which will receive the IFT CLI's command via an environment variable and allow you to get a check on whether it ran through or failed. The cylc workflow tests will be separate. There will be no makefile.
The tests of correctness of functionality should be in the Julia package.
…es' into jghrefactor/A2g-add-labeled-version-of-floes
Co-authored-by: Carlos Paniagua <[email protected]>
Co-authored-by: Carlos Paniagua <[email protected]>
Co-authored-by: Carlos Paniagua <[email protected]>
| imgs_ = [load_labeled_img(img) for img in imgs] | ||
| @info "Loading $props" | ||
| props_ = [DataFrame(CSV.File(prop)) for prop in props] | ||
| # go through each of the props_ dataframes and convert each | ||
| # into the element type from the corresponding image. | ||
| for (img_, prop_) in zip(imgs_, props_) | ||
| label_type = eltype(img_) | ||
| @debug "converting labels to $label_type" | ||
| prop_[!,:label] = convert.(label_type, prop_[!,:label]) | ||
| end |
There was a problem hiding this comment.
This suggests you might not be using a formatter. We've been using blue for Julia files.
# .JuliaFormatter.toml
style = "blue"
indent = 4There was a problem hiding this comment.
I've added this to #197
| uncasted_image = convert_uint_from_gray(casted_image) | ||
|
|
||
| @test isequal(image, uncasted_image) | ||
| @test isequal(eltype(image), eltype(uncasted_image)) |
There was a problem hiding this comment.
| @test isequal(eltype(image), eltype(uncasted_image)) | |
| @test isequal(T, eltype(uncasted_image)) |
There was a problem hiding this comment.
I don't think that change helps the reader understand the meaning of the test. The test as it is right now says "the eltypes of the images should be the same" whereas you're suggesting something which I'd read as "the eltype of the uncasted image should be the same as the element type of the argument to the function we're in" – which although being essentially the same thing, makes you have to think one level more abstractly when parsing the test.
Changes:
ToDo: