feat: expose CosemLabel as option in simulation (#18)

* cosem views

* wip

* wip

* wip

* wip

* style(pre-commit.ci): auto fixes [...]

* add download

* cleanup

* style(pre-commit.ci): auto fixes [...]

* better bin

* faster download

* working rebin

* style(pre-commit.ci): auto fixes [...]

* misc

* lots of stuff to run a cosem generation

* wip

* misc changes

* minor change

* update example

* update

* misc changes

* fix: remove caching on bins

* remove self type

* linting

* minor changes

* fallback to tight_layout

* rename cosem sample

* style(pre-commit.ci): auto fixes [...]

* changes

* pin griffe

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

Author: tlambert03

examples/basic_confocal.py

@@ -3,15 +3,17 @@
 from microsim import schema as ms
 from microsim.util import ndview, ortho_plot
 
-GreenMats = ms.FluorophoreDistribution(
-    distribution=ms.MatsLines(density=0.5, length=30, azimuth=5, max_r=1),
-    fluorophore="EGFP",
-)
-
 sim = ms.Simulation(
-    truth_space=ms.ShapeScaleSpace(shape=(128, 512, 512), scale=(0.02, 0.01, 0.01)),
-    output_space={"downscale": 8},
-    sample=ms.Sample(labels=[GreenMats]),
+    truth_space=ms.ShapeScaleSpace(shape=(64, 256, 256), scale=(0.04, 0.02, 0.02)),
+    output_space={"downscale": 4},
+    sample=ms.Sample(
+        labels=[
+            ms.FluorophoreDistribution(
+                distribution=ms.MatsLines(density=0.5, length=30, azimuth=5, max_r=1),
+                fluorophore="EGFP",
+            )
+        ]
+    ),
     modality=ms.Confocal(pinhole_au=0.5),
     settings=ms.Settings(random_seed=100, max_psf_radius_aus=8),
     detector=ms.CameraCCD(qe=0.82, read_noise=2, bit_depth=12),

examples/cosem.py

@@ -0,0 +1,30 @@
+from microsim import schema as ms
+from microsim.schema.optical_config import lib
+from microsim.util import ndview
+
+sim = ms.Simulation(
+    # note: this is a rather coarse simulation, but it's fast
+    # scale should be a one of .004 * 2^n, where n is an integer from 0 to 4
+    # space basically determines the field of view.
+    truth_space=ms.ShapeScaleSpace(shape=(52, 512, 512), scale=(0.064, 0.064, 0.064)),
+    output_space={"downscale": 2},
+    sample=ms.Sample(
+        labels=[
+            # pick dataset and layer name from https://openorganelle.janelia.org/datasets
+            ms.FluorophoreDistribution(
+                distribution=ms.CosemLabel(dataset="jrc_hela-3", label="er-mem_pred"),
+                fluorophore="EGFP",
+            ),
+            ms.FluorophoreDistribution(
+                distribution=ms.CosemLabel(dataset="jrc_hela-3", label="mito-mem_pred"),
+                fluorophore="mCherry",
+            ),
+        ]
+    ),
+    channels=[lib.FITC, lib.DSRED],
+    modality=ms.Widefield(),
+    detector=ms.CameraCCD(qe=0.82, read_noise=6),
+    settings=ms.Settings(max_psf_radius_aus=2),
+)
+
+ndview(sim.digital_image())

examples/confocal.ipynb examples/notebooks/confocal.ipynbexamples/confocal.ipynb renamed to examples/notebooks/confocal.ipynb

@@ -72,6 +72,7 @@ docs = [
     "mkdocs-material==9.5.17",
     "mkdocstrings ==0.24.3",
     "mkdocstrings-python ==1.9.2",
+    "griffe==0.45.3",
     "griffe-fieldz",
 ]
 

examples/notebooks/cosem.ipynb

@@ -71,7 +71,7 @@ def _supabase(url: str | None = None, key: str | None = None) -> supabase.Client
             text = response.read().decode("utf-8")
         key = text.split("SUPABASE_KEY:")[1].split(",")[0].strip("\"'")
         url = text.split("SUPABASE_URL:")[1].split(",")[0].strip("\"'")
-    return Client(url, key)  # type: ignore
+    return Client(url, key)
 
 
 @cache

examples/direct_truth.ipynb examples/notebooks/direct_truth.ipynbexamples/direct_truth.ipynb renamed to examples/notebooks/direct_truth.ipynb

@@ -262,12 +262,12 @@ def _sum_bin_chunk(
     dest_slice = tuple(slice(*x) for x in dest)
     block = in_arr[src_slice]
     # TODO: see if tensorstore has a better way to check if a block is empty
-    if np.any(block):  # type: ignore
+    if np.any(block):
         binned = block_bin(block.translate_to[0, 0, 0])
         if np.any(binned):
             # note, we could also use .write(), which is async...
             # but we're already in a thread, so it's not necessary?
-            out_arr[dest_slice] = binned.astype(out_arr.dtype.numpy_dtype)  # type: ignore
+            out_arr[dest_slice] = binned.astype(out_arr.dtype.numpy_dtype)
 
 
 def block_bin(
@@ -294,6 +294,6 @@ def block_bin(
     sliced = block[slc]
 
     # reshape the block and sum the values
-    reshaped_block = np.reshape(sliced, new_shape)  # type: ignore
+    reshaped_block = np.reshape(sliced, new_shape)
     axes = tuple(range(1, reshaped_block.ndim, 2))
     return ufunc(reshaped_block, axis=axes)

pyproject.toml

@@ -137,7 +137,7 @@ def create_dataarray(
     if attrs is None:
         attrs = dict(getattr(element, "attrs", {}))
     if use_dask:
-        element = da.from_array(element, chunks=chunks, inline_array=True)  # type: ignore
+        element = da.from_array(element, chunks=chunks, inline_array=True)
     result = xr.DataArray(element, coords=coords, attrs=attrs, name=name)
     return result
 

src/microsim/cosem/_client.py

@@ -97,7 +97,7 @@ class CosemImage(BaseModel):
     def bucket_path(self) -> str:
         return urllib.parse.urlparse(self.url).path
 
-    @computed_field
+    @computed_field  # type: ignore
     @property
     def scales(self) -> list[str]:
         """Fetch all available scales for the image from s3."""
@@ -155,6 +155,8 @@ def read(
         from microsim.cosem._tstore import read_tensorstore
 
         if bin_mode == "auto":
+            # we convert segmentation images to sum mode, because they are encoded
+            # for instance segmentation, which is not what we want.
             bin_mode = "sum" if self.content_type == "segmentation" else "standard"
         return read_tensorstore(
             self, level=level, transpose=transpose, bin_mode=bin_mode
@@ -293,7 +295,7 @@ def thumbnail(self) -> "np.ndarray":
                 "To use the thumbnail property, install the imageio package."
             ) from e
 
-        return imread(self.thumbnail_url)
+        return imread(self.thumbnail_url)  # type: ignore [no-any-return]
 
     def read(
         self, image_keys: str | Sequence[str], **read_kwargs: Any