resource efficiency update

CHANGELOG.md

@@ -3,6 +3,22 @@
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/)
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+
+## v0.1.1 - [2025.11.15]
+
+Efficiency update - only cells with overlaps are taken into consideration when calculating IoU matrices. Mask comparisons not calculated through a direct comparison, but optimized through lookup tables.
+
+### `Added`
+- `plot_target_size` parameter for plotting - inputs will be "downscaled" to match the target size for plotting efficiency
+
+### `Fixed`
+- memory issues for large input masks with many cells
+- Dockerfile copies the repo to pip install instead of pointing to main
+
+### `Removed`
+- graph construction with networkx, replaced with functions
+- cost matrix metric currently doesn't affect the execution - only IoU is used. To be updated soon.
+
 ## v0.1.0 - [2025.11.12]
 
 First release of Segobe - a tool for object matching, segmentation error counting and metric evaluation.

Dockerfile

@@ -18,3 +18,8 @@ USER root
 # Ensure micromamba binaries are in PATH
 ENV PATH="$MAMBA_ROOT_PREFIX/bin:$PATH"
 
+# Copy the rest of the current directory into /app inside the container
+WORKDIR /app
+COPY . .
+
+RUN micromamba run -n base pip install .

README.md

@@ -8,7 +8,7 @@ Designed for cell segmentation evaluation, it can handle large batches of sample
 
 ## Installation
 
-Option 1. Install directory from the repository (recommended for development)
+### Option 1. Install directory from the repository (recommended for development)
 If you plan to develop or modify Segobe, install it in editable mode:
 ```bash
 # Clone the repository
@@ -24,7 +24,7 @@ pip install -e .
 ```
 > The -e flag (editable mode) means the changes to the source code are immediately reflected without reinstalling.
 
-Option 2. Install directly from GitHub
+### Option 2. Install directly from GitHub
 Once the repository is made public, users can install it directly via URL:
 ```bash
 pip install git+https://github.com/schapirolabor/segobe.git
@@ -50,7 +50,6 @@ segobe \
     --iou_threshold 0.5 \
     --graph_iou_threshold 0.1 \
     --unmatched_cost 0.4 \
-    --cost_matrix_metric 'iou' \
     --save_plots
 ```
 
@@ -62,8 +61,9 @@ segobe \
 |  | --iou_threshold | IoU threshold for cell matching (0-1, default: 0.5). Match is true if pair is selected with linear_sum_assignment and IoU above this threshold. |
 |  | --graph_iou_threshold | Graph IoU threshold for error detection (0-1, default: 0.1). Minimal IoU for cells to be considered 'connected'. |
 |  | --unmatched_cost | Cost for unmatched objects in the cost matrix (0-1, default: 0.4) |
-|  | --cost_matrix_metric | Specify which metric should be used for cost matrix construction (default: 'iou', other options 'dice', 'moc' - see details [here](docs/detailed_overview.md)) |
+|  | --cost_matrix_metric | Specify which metric should be used for cost matrix construction (default: 'iou', other options 'dice', 'moc' - see details [here](docs/detailed_overview.md)) `note that only IoU is currently supported` |
 |  | --save_plots | Boolean specifying whether plots (barplot grouped by category and row-specific error overview) are saved |
+|  | --plot_target_size | Size in pixels of the plot error types subfigures. If the inputs are larger, they will be approximately downsampled by a scale factor. If that scale factor is larger than 4, boundaries will not be drawn. (default: 600) |
 |  | --version | Prints tool version. |
 
 ### Input format
@@ -110,7 +110,6 @@ Captured metrics in the `metrics.csv` for each input CSV `row`:
 * Splits: counts and dictionary of matched predictions to GTs
 * Merges: counts and dictionary of matched GTs to predictions
 * Catastrophes: counts and dictionary of groups of GTs and predictions involved
-* IoU graph: constructed graph of cell overlaps
 * True postives: counts
 * False positives: counts
 * False negatives: counts

environment.yml

@@ -9,7 +9,6 @@ dependencies:
   - scipy
   - scikit-image
   - matplotlib
-  - networkx
   - tifffile
   - git
   - pip

pyproject.toml

@@ -19,7 +19,6 @@ dependencies = [
     "pandas",
     "matplotlib",
     "scipy",
-    "networkx",
     "scikit-image",
     "tifffile"
 ]

segobe/CLI.py

@@ -59,6 +59,12 @@ def get_args():
         help="Metric used for cost matrix calculation (default: iou)",
         choices=["iou", "dice", "moc"],
     )
+    parser.add_argument(
+        "--target_plot_size",
+        type=int,
+        default=600,
+        help="Target size to which large input images will be downsampled for error type plots.",
+    )
     parser.add_argument(
         "--save_plots",
         action="store_true",
@@ -84,8 +90,12 @@ def main():
 
     # Read input CSV
     df = pd.read_csv(args.input_csv)
+    required_columns = {"sampleID", "ref_mask", "eval_mask", "category"}
+    if not required_columns.issubset(df.columns):
+        raise ValueError(
+            f"Input CSV must contain columns: {required_columns}, got {df.columns}"
+        )
 
-    print(args.cost_matrix_metric)
     # Run batch evaluation
     batch_eval = SegmentationEvaluationBatch(
         df,
@@ -136,6 +146,8 @@ def main():
                     f"{args.basename}_{row['sampleID']}_{row['category']}_error_types.png",
                 ),
                 suptitle=True,
+                legend=False,
+                target_size=args.target_plot_size
             )
 
 

segobe/__init__.py

@@ -2,7 +2,7 @@
 from .plotter import plot_error_types, plot_barplot
 from .utils import filter_mask_by_ids
 
-__version__ = "0.1.0"
+__version__ = "0.1.1"
 __all__ = [
     "SegmentationEvaluator",
     "SegmentationEvaluationBatch",