Merge branch 'master' into dev

Author: anwai98

RELEASE_OVERVIEW.md

@@ -1,5 +1,9 @@
 # Release Overview
 
+**New in version 1.7.1**
+
+Fixing minor issues in 1.7.0 (related to trackastra, automatic segmentation and training functions) and adding new section in documentation for our new automatic segmentation pipeline, APG.
+
 **New in version 1.7.0**
 
 Updates to the automatic instance segmentation pipeline (introduces APG - automatic prompt generation).

doc/deprecated/development.md

@@ -5,7 +5,7 @@
 This software consists of four different python (sub-)modules:
 - The top-level `micro_sam` module implements general purpose functionality for using Segment Anything for multi-dimensional data.
 - `micro_sam.evaluation` provides functionality to evaluate Segment Anything models on (microscopy) segmentation tasks.
-- `micro_sam.traning` implements the training functionality to finetune Segment Anything for custom segmentation datasets.
+- `micro_sam.training` implements the training functionality to finetune Segment Anything for custom segmentation datasets.
 - `micro_sam.sam_annotator` implements the interactive annotation tools.
 
 ## Annotation Tools

examples/object_classifier.py

@@ -82,7 +82,7 @@ def _get_livecell_data():
 
     embedding_path = os.path.join(EMBEDDING_CACHE, "embeddings-livecell-vit_b_lm.zarr")
 
-    # This is the vit-b-lm segmentation and a test annotaiton.
+    # This is the vit-b-lm segmentation and a test annotation.
     segmentation = imageio.imread("./clf-test-data/livecell-test-seg.tif")
     annotations = imageio.imread("./clf-test-data/livecell-test-annotations.tif")
 
@@ -97,7 +97,7 @@ def _get_wholeslide_data():
 
     embedding_path = os.path.join(EMBEDDING_CACHE, "whole-slide-embeddings-vit_b_lm.zarr")
 
-    # This is the vit-b-lm segmentation and a test annotaiton.
+    # This is the vit-b-lm segmentation and a test annotation.
     segmentation = imageio.imread("./clf-test-data/whole-slide-seg.tif")
     annotations = imageio.imread("./clf-test-data/wholeslide-annotations.tif")
 

micro_sam/__version__.py

@@ -1 +1 @@
-__version__ = "1.7.0"
+__version__ = "1.7.1"

micro_sam/bioimageio/model_export.py

@@ -283,7 +283,7 @@ def export_sam_model(
 
     Args:
         image: The image for generating test data.
-        label_image: The segmentation correspoding to `image`.
+        label_image: The segmentation corresponding to `image`.
             It is used to derive prompt inputs for the model.
         model_type: The type of the SAM model.
         name: The name of the exported model.

micro_sam/instance_segmentation.py

@@ -1408,7 +1408,7 @@ def prompt_function(foreground, center_distances, boundary_distances, **kwargs)
 
         # 2.) Apply the predictor to the prompts.
         if prompts is None:  # No prompts were derived, we can't do much further and return empty masks.
-            return np.zeros(foreground.shape, dtype="uint32") if output_mode == "instance_egmentation" else []
+            return np.zeros(foreground.shape, dtype="uint32") if output_mode == "instance_segmentation" else []
         else:
             predictions = batched_inference(
                 self._predictor,

micro_sam/models/peft_sam.py

@@ -269,7 +269,7 @@ def allow_gradient_update_for_parameters(
         Args:
             prefix: Matches the part of parameter name in front.
             suffix: Matches the part of parameter name at the end.
-            infix: Matches parts of parameter name occuring in between.
+            infix: Matches parts of parameter name occurring in between.
         """
         for k, v in self.block.named_parameters():
             if prefix is not None and k.startswith(tuple(prefix)):

micro_sam/multi_dimensional_segmentation.py

@@ -567,7 +567,7 @@ def _filter_lineages(lineages, tracking_result):
 def _tracking_impl(timeseries, segmentation, mode, min_time_extent, output_folder=None):
     device = "cuda" if torch.cuda.is_available() else "cpu"
     model = Trackastra.from_pretrained("general_2d", device=device)
-    lineage_graph = model.track(timeseries, segmentation, mode=mode)
+    lineage_graph, _ = model.track(timeseries, segmentation, mode=mode)
     track_data, parent_graph, _ = graph_to_napari_tracks(lineage_graph)
     node_to_track, lineages = _extract_tracks_and_lineages(segmentation, track_data, parent_graph)
     tracking_result = recolor_segmentation(segmentation, node_to_track)

micro_sam/sam_annotator/util.py

@@ -448,7 +448,7 @@ def _batched_interactive_segmentation(predictor, points, labels, boxes, image_em
     # (This is left here as a reference for how this can be implemented.
     #  I have not decided yet if this is actually a good idea or not.)
     # # If we have no objects: this is the first call for a batched segmentation.
-    # # We treat each positive point or box as a separate obejct.
+    # # We treat each positive point or box as a separate object.
     # if len(seg_ids) == 1:
     #     # Create a list of all prompts.
     #     batched_prompts = [(None, point, label) for point, label in zip(batched_points, batched_labels)]

micro_sam/util.py

@@ -1360,7 +1360,7 @@ def segmentation_to_one_hot(segmentation: np.ndarray, segmentation_ids: Optional
     masks = masks.unsqueeze(0)  # add dimension to scatter
     masks = torch.zeros(one_hot_shape).scatter_(0, masks, 1)[1:]
 
-    # add the extra singleton dimenion to get shape NUM_OBJECTS x 1 x H x W
+    # add the extra singleton dimension to get shape NUM_OBJECTS x 1 x H x W
     masks = masks.unsqueeze(1)
     return masks