KappaIOU 1.0.0__tar.gz

kappaiou-1.0.0/KappaIOU.egg-info/PKG-INFO

@@ -0,0 +1,41 @@
+Metadata-Version: 2.4
+Name: KappaIOU
+Version: 1.0.0
+Summary: A training-free, chance-corrected reliability library for continuous multi-object spatial annotations.
+Author-email: "F. Trevor Rogers" <rogers72@hawaii.edu>
+Project-URL: Homepage, https://github.com/tr7200/KappaIOU
+Project-URL: Bug Tracker, https://github.com/tr7200/KappaIOU/issues
+Classifier: Programming Language :: Python :: 3
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: numpy>=1.20.0
+Requires-Dist: scipy>=1.7.0
+Dynamic: license-file
+
+# Kappa-IoU
+
+This Python package implements the Kappa-IoU measure of Inter-annotator reliability from Rogers (2026):
+
+    Rogers, Trevor F. (2026). Kappa-IoU: Inter-Rater Reliability for Spatial Annotation
+       
+Large-scale annotation projects for which automated annotation solutions are insufficient require a metric that measures the quality of annotation between annotators. 
+[Intersection-over-Union](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=Rezatofighi%2C%20H.%2C%20Tsoi%2C%20N.%2C%20Gwak%2C%20J.%2C%20Sadeghian%2C%20A.%2C%20Reid%2C%20I.%2C%20%26%20Savarese%2C%20S.%20(2019).%20Generalized%20intersection%20over%20union%3A%20A%20metric%20and%20a%20loss%20for%20object%20detection.%20Proceedings%20of%20the%20IEEE%2FCVF%20Conference%20on%20Computer%20Vision%20and%20Pattern%20Recognition%20(CVPR).), or IoU, is traditionally used for spatial data but requires a ground truth. 
+[Cohen's Kappa](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=Cohen%2C%20J.%20(1960).%20A%20coefficient%20of%20agreement%20for%20nominal%20scales.%C2%A0Educational%20and%20psychological%20measurement%2C%C2%A020(1)%2C%2037-46.) is a psychometric measure that measures the amount of agreement between raters, but is traditionally used for categorical data.
+
+This repository contains a Python package implementing the [Kappa-IoU](https://drive.google.com/file/d/1QW1dbFWX1-rxz7XcvPkBdlR0BFKoBIwC/view?usp=sharing) paper, which indexes IoU with Cohen's Kappa, and handles differences in object cardinality between raters by treating it as a linear assignment problem using [Kuhn-Munkres](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=Munkres%2C%20J.%20(1957).%20Algorithms%20for%20the%20assignment%20and%20transportation%20problems.%20Journal%20of%20the%20Society%20for%20Industrial%20and%20Applied%20Mathematics%2C%205(1)%2C%2032-38.) assignment. Also included is an adjusted version of Kappa-IoU that uses a version of the [BLEU](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=Papineni%2C%20K.%2C%20Roukos%2C%20S.%2C%20Ward%2C%20T.%2C%20%26%20Zhu%2C%20W.%20J.%20(2002).%20BLEU%3A%20a%20method%20for%20automatic%20evaluation%20of%20machine%20translation.%20In%20Proceedings%20of%20the%2040th%20Annual%20Meeting%20of%20the%20Association%20for%20Computational%20Linguistics%20(pp.%20311-318).) measure for machine translation which instead scores N-objects for multi-object annotation. Along with the Python package implementing this measure, this repository also contains `Section4.py`, code that recreates the examples from that section of the paper.
+
+To install this package from inside the downloaded folder from Github:
+
+    python -m pip install -e .
+
+To use, run the analysis engine straight from your terminal shell across your dataset directories using the console entrypoint:
+
+    kappaiou-run --baseline /data/author_baseline.csv --dir /data/annotator_tranches/ --tau 0.70
+
+The author baseline and annotator tranche CSV files are expected to have the filenames in the first column, label in the second column, and then xywh-format annotations in the following four columns.
+
+MIT License

kappaiou-1.0.0/KappaIOU.egg-info/SOURCES.txt

@@ -0,0 +1,13 @@
+LICENSE
+README.md
+pyproject.toml
+KappaIOU.egg-info/PKG-INFO
+KappaIOU.egg-info/SOURCES.txt
+KappaIOU.egg-info/dependency_links.txt
+KappaIOU.egg-info/entry_points.txt
+KappaIOU.egg-info/requires.txt
+KappaIOU.egg-info/top_level.txt
+kappaiou/__init__.py
+kappaiou/cli.py
+kappaiou/core.py
+tests/test.py

kappaiou-1.0.0/KappaIOU.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

kappaiou-1.0.0/KappaIOU.egg-info/entry_points.txt

@@ -0,0 +1,2 @@
+[console_scripts]
+kappaiou-run = kappaiou.cli:main

kappaiou-1.0.0/KappaIOU.egg-info/requires.txt

@@ -0,0 +1,2 @@
+numpy>=1.20.0
+scipy>=1.7.0

kappaiou-1.0.0/KappaIOU.egg-info/top_level.txt

@@ -0,0 +1 @@
+kappaiou

kappaiou-1.0.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 T. Rogers
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

kappaiou-1.0.0/PKG-INFO

@@ -0,0 +1,41 @@
+Metadata-Version: 2.4
+Name: KappaIOU
+Version: 1.0.0
+Summary: A training-free, chance-corrected reliability library for continuous multi-object spatial annotations.
+Author-email: "F. Trevor Rogers" <rogers72@hawaii.edu>
+Project-URL: Homepage, https://github.com/tr7200/KappaIOU
+Project-URL: Bug Tracker, https://github.com/tr7200/KappaIOU/issues
+Classifier: Programming Language :: Python :: 3
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: numpy>=1.20.0
+Requires-Dist: scipy>=1.7.0
+Dynamic: license-file
+
+# Kappa-IoU
+
+This Python package implements the Kappa-IoU measure of Inter-annotator reliability from Rogers (2026):
+
+    Rogers, Trevor F. (2026). Kappa-IoU: Inter-Rater Reliability for Spatial Annotation
+       
+Large-scale annotation projects for which automated annotation solutions are insufficient require a metric that measures the quality of annotation between annotators. 
+[Intersection-over-Union](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=Rezatofighi%2C%20H.%2C%20Tsoi%2C%20N.%2C%20Gwak%2C%20J.%2C%20Sadeghian%2C%20A.%2C%20Reid%2C%20I.%2C%20%26%20Savarese%2C%20S.%20(2019).%20Generalized%20intersection%20over%20union%3A%20A%20metric%20and%20a%20loss%20for%20object%20detection.%20Proceedings%20of%20the%20IEEE%2FCVF%20Conference%20on%20Computer%20Vision%20and%20Pattern%20Recognition%20(CVPR).), or IoU, is traditionally used for spatial data but requires a ground truth. 
+[Cohen's Kappa](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=Cohen%2C%20J.%20(1960).%20A%20coefficient%20of%20agreement%20for%20nominal%20scales.%C2%A0Educational%20and%20psychological%20measurement%2C%C2%A020(1)%2C%2037-46.) is a psychometric measure that measures the amount of agreement between raters, but is traditionally used for categorical data.
+
+This repository contains a Python package implementing the [Kappa-IoU](https://drive.google.com/file/d/1QW1dbFWX1-rxz7XcvPkBdlR0BFKoBIwC/view?usp=sharing) paper, which indexes IoU with Cohen's Kappa, and handles differences in object cardinality between raters by treating it as a linear assignment problem using [Kuhn-Munkres](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=Munkres%2C%20J.%20(1957).%20Algorithms%20for%20the%20assignment%20and%20transportation%20problems.%20Journal%20of%20the%20Society%20for%20Industrial%20and%20Applied%20Mathematics%2C%205(1)%2C%2032-38.) assignment. Also included is an adjusted version of Kappa-IoU that uses a version of the [BLEU](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=Papineni%2C%20K.%2C%20Roukos%2C%20S.%2C%20Ward%2C%20T.%2C%20%26%20Zhu%2C%20W.%20J.%20(2002).%20BLEU%3A%20a%20method%20for%20automatic%20evaluation%20of%20machine%20translation.%20In%20Proceedings%20of%20the%2040th%20Annual%20Meeting%20of%20the%20Association%20for%20Computational%20Linguistics%20(pp.%20311-318).) measure for machine translation which instead scores N-objects for multi-object annotation. Along with the Python package implementing this measure, this repository also contains `Section4.py`, code that recreates the examples from that section of the paper.
+
+To install this package from inside the downloaded folder from Github:
+
+    python -m pip install -e .
+
+To use, run the analysis engine straight from your terminal shell across your dataset directories using the console entrypoint:
+
+    kappaiou-run --baseline /data/author_baseline.csv --dir /data/annotator_tranches/ --tau 0.70
+
+The author baseline and annotator tranche CSV files are expected to have the filenames in the first column, label in the second column, and then xywh-format annotations in the following four columns.
+
+MIT License

kappaiou-1.0.0/README.md

@@ -0,0 +1,23 @@
+# Kappa-IoU
+
+This Python package implements the Kappa-IoU measure of Inter-annotator reliability from Rogers (2026):
+
+    Rogers, Trevor F. (2026). Kappa-IoU: Inter-Rater Reliability for Spatial Annotation
+       
+Large-scale annotation projects for which automated annotation solutions are insufficient require a metric that measures the quality of annotation between annotators. 
+[Intersection-over-Union](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=Rezatofighi%2C%20H.%2C%20Tsoi%2C%20N.%2C%20Gwak%2C%20J.%2C%20Sadeghian%2C%20A.%2C%20Reid%2C%20I.%2C%20%26%20Savarese%2C%20S.%20(2019).%20Generalized%20intersection%20over%20union%3A%20A%20metric%20and%20a%20loss%20for%20object%20detection.%20Proceedings%20of%20the%20IEEE%2FCVF%20Conference%20on%20Computer%20Vision%20and%20Pattern%20Recognition%20(CVPR).), or IoU, is traditionally used for spatial data but requires a ground truth. 
+[Cohen's Kappa](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=Cohen%2C%20J.%20(1960).%20A%20coefficient%20of%20agreement%20for%20nominal%20scales.%C2%A0Educational%20and%20psychological%20measurement%2C%C2%A020(1)%2C%2037-46.) is a psychometric measure that measures the amount of agreement between raters, but is traditionally used for categorical data.
+
+This repository contains a Python package implementing the [Kappa-IoU](https://drive.google.com/file/d/1QW1dbFWX1-rxz7XcvPkBdlR0BFKoBIwC/view?usp=sharing) paper, which indexes IoU with Cohen's Kappa, and handles differences in object cardinality between raters by treating it as a linear assignment problem using [Kuhn-Munkres](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=Munkres%2C%20J.%20(1957).%20Algorithms%20for%20the%20assignment%20and%20transportation%20problems.%20Journal%20of%20the%20Society%20for%20Industrial%20and%20Applied%20Mathematics%2C%205(1)%2C%2032-38.) assignment. Also included is an adjusted version of Kappa-IoU that uses a version of the [BLEU](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=Papineni%2C%20K.%2C%20Roukos%2C%20S.%2C%20Ward%2C%20T.%2C%20%26%20Zhu%2C%20W.%20J.%20(2002).%20BLEU%3A%20a%20method%20for%20automatic%20evaluation%20of%20machine%20translation.%20In%20Proceedings%20of%20the%2040th%20Annual%20Meeting%20of%20the%20Association%20for%20Computational%20Linguistics%20(pp.%20311-318).) measure for machine translation which instead scores N-objects for multi-object annotation. Along with the Python package implementing this measure, this repository also contains `Section4.py`, code that recreates the examples from that section of the paper.
+
+To install this package from inside the downloaded folder from Github:
+
+    python -m pip install -e .
+
+To use, run the analysis engine straight from your terminal shell across your dataset directories using the console entrypoint:
+
+    kappaiou-run --baseline /data/author_baseline.csv --dir /data/annotator_tranches/ --tau 0.70
+
+The author baseline and annotator tranche CSV files are expected to have the filenames in the first column, label in the second column, and then xywh-format annotations in the following four columns.
+
+MIT License

kappaiou-1.0.0/kappaiou/__init__.py

@@ -0,0 +1,7 @@
+from .core import calculate_iou, compute_image_kappa_iou, calculate_adjusted_kappa_iou
+
+__all__ = [
+    'calculate_iou',
+    'compute_image_kappa_iou',
+    'calculate_adjusted_kappa_iou'
+]

kappaiou-1.0.0/kappaiou/cli.py

@@ -0,0 +1,125 @@
+"""
+
+cli.py
+
+backend input parsing
+
+"""
+
+import os
+import sys
+import argparse
+import csv
+from collections import defaultdict
+from .core import calculate_adjusted_kappa_iou
+
+def parse_csv_annotations(file_path):
+    """
+    Parses your standard input structure:
+    Col 1: Filename | Col 2: Label (Object Type) | Col 3-6+: Spatial Data [x, y, w, h]
+
+    Input:
+    - file_path (str): path to annotations file
+
+    Returns:
+    - data (defaultdict): parsed annotations file
+    """
+    data = defaultdict(lambda: defaultdict(list))
+  
+    if not os.path.exists(file_path):
+        print(f"Error: Target file not found at {file_path}")
+        sys.exit(1)
+        
+    with open(file_path, mode='r', newline='', encoding='utf-8') as f:
+        reader = csv.reader(f)
+        # Skip optional headers if detected via string characters
+        first_row = next(reader, None)
+        
+        if first_row:
+            try:
+                float(first_row[2])
+                row_start = first_row
+            except ValueError:
+                row_start = None  # Row was a string header
+                
+            def _process_row(row):
+                if len(row) < 6:
+                    return
+                
+                filename = row[0]
+                label = row[1]
+              
+                try:
+                    coords = [float(c) for c in row[2:6]]
+                    data[filename][label].append(coords)
+                except ValueError:
+                    pass # Row logging exception safety
+                    
+            if row_start:
+                _process_row(row_start)
+            for row in reader:
+                _process_row(row)
+              
+    return data
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Run Kappa-IoU verification against downstream annotator tranches.")
+    parser.add_argument("--baseline",
+                        required=True,
+                        help="Path to your primary baseline CSV file.")
+    parser.add_argument("--dir",
+                        required=True,
+                        help="Directory containing all other annotator CSV files.")
+    parser.add_argument("--tau",
+                        type=float,
+                        default=0.70,
+                        help="Expected spatial project acceptance threshold.")
+    
+    args = parser.parse_args()
+    
+    # Parse Master/Baseline Dataset Template
+    baseline_data = parse_csv_annotations(args.baseline)
+    
+    # Gather other target CSV logs in target path directory
+    all_files = os.listdir(args.dir)
+    annotator_files = [f for f in all_files if f.endswith('.csv') and os.path.abspath(os.path.join(args.dir, f)) != os.path.abspath(args.baseline)]
+    
+    if not annotator_files:
+        print(f"No valid secondary annotator CSV files located in target directory: {args.dir}")
+        return
+
+    print(f"=========================================================================")
+    print(f"KAPPA-IoU COMPLIANCE LAB PIPELINE RUNTIME ENGINE")
+    print(f"=========================================================================")
+    print(f"Master baseline:       {os.path.basename(args.baseline)}")
+    print(f"Operational tolerance (\u03c4): {args.tau}")
+    print(f"Secondary systems found: {len(annotator_files)}")
+    print(f"-------------------------------------------------------------------------")
+
+    # Evaluate each annotator file independently
+    for target in annotator_files:
+        target_path = os.path.join(args.dir, target)
+        annotator_data = parse_csv_annotations(target_path)
+        
+        cumulative_score = 0.0
+        evaluation_channels = 0
+        
+        # Iterate cross-wise through every shared baseline domain image coordinate
+        for filename, labels_dict in baseline_data.items():
+            for label, base_boxes in labels_dict.items():
+                # Extract corresponding logs from target annotator
+                annotator_boxes = annotator_data.get(filename, {}).get(label, [])
+                
+                score = calculate_adjusted_kappa_iou(base_boxes, annotator_boxes, tau=args.tau)
+                cumulative_score += score
+                evaluation_channels += 1
+                
+        if evaluation_channels > 0:
+            final_metric = cumulative_score / evaluation_channels
+            print(f"Annotator File: {target:<30} | Adjusted Kappa-IoU: {final_metric:.4f}")
+        else:
+            print(f"Annotator File: {target:<30} | Adjusted Kappa-IoU: ERR (Zero matching image/label intersections)")
+
+if __name__ == "__main__":
+    main()

kappaiou-1.0.0/kappaiou/core.py

@@ -0,0 +1,134 @@
+"""
+
+Mathematical backend
+
+- IoU
+- Kappa-IoU
+- adjusted Kappa-IoU (using modified BLEU for N-objects)
+
+"""
+
+
+import numpy as np
+from scipy.optimize import linear_sum_assignment
+
+
+def calculate_iou(boxA, boxB):
+    """
+    Computes Intersection-over-Union (IoU) between two boxes in [xmin, ymin, w, h] format.
+
+    Input:
+    - boxA (list): xywh-format annotation
+    - boxB (list): xywh-format annotation
+
+    Returns:
+    - Intersection-over-Union of boxA and boxB
+    """
+    xA = max(boxA[0], boxB[0])
+    yA = max(boxA[1], boxB[1])
+    xB = min(boxA[0] + boxA[2], boxB[0] + boxB[2])
+    yB = min(boxA[1] + boxA[3], boxB[1] + boxB[3])
+    
+    intersection_area = max(0, xB - xA) * max(0, yB - yA)
+    boxA_area = boxA[2] * boxA[3]
+    boxB_area = boxB[2] * boxB[3]
+    union_area = boxA_area + boxB_area - intersection_area
+    
+    if union_area == 0:
+        return 0.0
+    
+    return intersection_area / union_area
+
+
+def compute_image_kappa_iou(rater_a_boxes, 
+                            rater_b_boxes, 
+                            tau=0.70, 
+                            delta=1e-7):
+    """
+    In case of raters annotating different numbers of objects on an image, 
+    this difference is solved as a linear assignment problem via Kuhn-Munkres 
+    and Kappa-IoU is calculated for a single object-set channel.
+    
+    Input:
+    - rater_a_boxes (List): List of lists of boxes annotated by rater A
+    - rater_b_boxes (List): List of lists of boxes annotated by rater B
+    - tau (float): P_T, the acceptable threshold of agreement between rater A and rater B
+    - delta (float): stabilizing constant to prevent division by zero
+
+    Returns:
+    - Kappa-IoU metric of rater A and rater B
+    """
+    N_A = len(rater_a_boxes)
+    N_B = len(rater_b_boxes)
+    max_N = max(N_A, N_B)
+    
+    if max_N == 0:
+        return 1.0  # Perfect consensus results on an empty image channel
+        
+    # Build cost matrix (1 - IoU)
+    cost_matrix = np.zeros((N_A, N_B))
+    for j in range(N_A):
+        for k in range(N_B):
+            cost_matrix[j, k] = 1.0 - calculate_iou(rater_a_boxes[j], rater_b_boxes[k])
+            
+    # Pad to square configuration with max cost penalty of 1.0
+    padded_cost = np.ones((max_N, max_N))
+    padded_cost[:N_A, :N_B] = cost_matrix
+    
+    # Primal-dual optimization path routing
+    row_ind, col_ind = linear_sum_assignment(padded_cost)
+    
+    optimized_ious = []
+    for r, c in zip(row_ind, col_ind):
+        if r >= N_A or c >= N_B:
+            optimized_ious.append(0.0)  # Unmatched dummy penalty
+        else:
+            optimized_ious.append(1.0 - cost_matrix[r, c])
+            
+    avg_iou = np.mean(optimized_ious)
+                              
+    P_0 = 1.0 / (1.0 - avg_iou + delta)
+    kappaiou = (P_0 - tau) / (1.0 - tau)
+                              
+    return kappaiou
+
+
+def calculate_adjusted_kappa_iou(rater_a_boxes, 
+                                 rater_b_boxes, 
+                                 tau=0.70, 
+                                 delta=1e-7):
+    """
+    Implements the BLEU-adapted geometric ranking summation framework.
+    Evaluates sub-sequences from 1 to N-objects and aggregates the reliability index.
+    
+    Input:
+    - rater_a_boxes (List): List of lists of boxes annotated by rater A
+    - rater_b_boxes (List): List of lists of boxes annotated by rater B
+    - tau (float): P_T, the acceptable threshold of agreement between rater A and rater B
+    - delta (float): stabilizing constant to prevent division by zero
+
+    Returns:
+    - Kappa-IoU metric of rater A and rater B adjusted by modified BLEU
+    """
+    N_A = len(rater_a_boxes)
+    N_B = len(rater_b_boxes)
+    N = max(N_A, N_B)
+    
+    if N == 0:
+        return 0.0
+        
+    kappa_terms = []
+                                   
+    # Evaluate sequentially across every depth layer from 1 to N
+    for n in range(1, N + 1):
+        sub_a = rater_a_boxes[:n]
+        sub_b = rater_b_boxes[:n]
+        k_n = compute_image_kappa_iou(sub_a, sub_b, tau=tau, delta=delta)
+        kappa_terms.append(k_n)
+        
+    # Structural geometric log-precision summation sequence
+    # Protects against taking natural logs of zero or negative Kappa outputs
+    valid_logs = [np.log(k) if k > 0 else np.log(delta) for k in kappa_terms]
+    adjusted_score = np.exp(np.sum((1.0 / N) * np.array(valid_logs)))
+                                   
+    return adjusted_score

kappaiou-1.0.0/pyproject.toml

@@ -0,0 +1,41 @@
+[build-system]
+requires = ["setuptools>=61.0.0", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "KappaIOU"
+version = "1.0.0"
+authors = [
+    { name="F. Trevor Rogers", email="rogers72@hawaii.edu" }
+]
+description = "A training-free, chance-corrected reliability library for continuous multi-object spatial annotations."
+readme = "README.md"
+requires-python = ">=3.8"
+classifiers = [
+    "Programming Language :: Python :: 3",
+    "License :: OSI Approved :: MIT License",
+    "Operating System :: OS Independent",
+    "Topic :: Scientific/Engineering :: Artificial Intelligence",
+]
+dependencies = [
+    "numpy>=1.20.0",
+    "scipy>=1.7.0",
+]
+
+[project.urls]
+"Homepage" = "https://github.com/tr7200/KappaIOU"
+"Bug Tracker" = "https://github.com/tr7200/KappaIOU/issues"
+
+[project.scripts]
+kappaiou-run = "kappaiou.cli:main"
+
+[tool.setuptools.packages.find]
+where = ["."]
+include = ["kappaiou*"]
+
+[tool.pytest.ini_options]
+minversion = "6.0"
+testpaths = ["tests"]
+
+[tool.ruff]
+line-length = 88

kappaiou-1.0.0/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+

kappaiou-1.0.0/tests/test.py

@@ -0,0 +1,98 @@
+"""
+
+Unit tests
+
+"""
+
+import pytest
+import numpy as np
+
+from kappaiou.core import (
+    calculate_iou, 
+    compute_image_kappa_iou, 
+    calculate_adjusted_kappa_iou
+)
+
+# =========================================================================
+# EMPIRICAL TRUNCATED RAW DATA CONSTANTS (FROM MANUSCRIPT SECTION 4)
+# =========================================================================
+
+# Tranche 1: Single-Object (Cat)
+T1_RATER_A = [[2.7439, 53.506, 373.6281, 273.018]]
+T1_RATER_B = [[75.457, 65.854, 418.903, 257.469]]
+
+# Tranche 2: Multi-Object Inverted Sequence (Cows)
+T2_RATER_A = [
+    [96.951, 105.64024, 156.403, 166.92076],  # Left Cow
+    [246.494, 169.665, 143.14015, 105.18256]   # Right Cow
+]
+T2_RATER_B = [
+    [244.207, 173.323, 148.171, 101.982],      # Right Cow (Inverted)
+    [94.207, 104.726, 160.061, 167.378]        # Left Cow (Inverted)
+]
+
+# Tranche 3: Multi-Object Missingness/Symmetric Execution (Dogs)
+T3_RATER_A = [
+    [92.835, 163.72, 126.22, 134.451],         # Left Running Dog
+    [252.439, 194.36, 109.756, 80.488],        # Center Spotted Dog
+    [325.61, 195.732, 91.463, 89.177]          # Right Spotted Dog
+]
+T3_RATER_B = [
+    [97.5, 168.0, 123.5, 125.0],               # Left Running Dog
+    [322.5, 190.5, 93.0, 94.5],                # Right Spotted Dog (Inverted)
+    [254.5, 196.5, 109.5, 83.5]                # Center Spotted Dog (Inverted)
+]
+
+# =========================================================================
+# PYTEST COMPLIANCE LAB TEST SUITE
+# =========================================================================
+
+def test_tranche_1_single_object_scaling():
+    """
+    Verifies that Tranche 1 (Cat) properly exhibits the open-ended positive
+    scaling characteristics when P_0 exceeds classical limits.
+    """
+    tau = 0.70
+    delta = 1e-7
+    
+    # Extract structural baseline behavior
+    kappa_score = compute_image_kappa_iou(T1_RATER_A, T1_RATER_B, tau=tau, delta=delta)
+    
+    # Assertions grounded in Section 4 empirical findings
+    assert kappa_score > 1.0, f"Expected Kappa to break upper bound, got {kappa_score}"
+    assert np.isfinite(kappa_score), "Kappa calculation returned unstable non-finite float."
+
+
+def test_tranche_2_sequence_invariance():
+    """
+    Verifies that the Kuhn-Munkres engine unscrambles arbitrary sequence inversions,
+    yielding a highly stable score regardless of rater entry order.
+    """
+    tau = 0.70
+    delta = 1e-7
+    
+    # Run structural assessment on the cows (inverted indexes)
+    kappa_score = compute_image_kappa_iou(T2_RATER_A, T2_RATER_B, tau=tau, delta=delta)
+    
+    # Construct a perfectly sequential reference to verify invariance
+    sequenced_rater_b = [T2_RATER_B[1], T2_RATER_B[0]]
+    reference_score = compute_image_kappa_iou(T2_RATER_A, sequenced_rater_b, tau=tau, delta=delta)
+    
+    # Assert that the optimization resolves order variance completely
+    assert kappa_score == pytest.approx(reference_score, rel=1e-6)
+
+
+def test_tranche_3_adjusted_sequential_bleu():
+    """
+    Verifies that the multi-object BLEU aggregation accurately combines 
+    sub-sequence Kappa scores across varying depths up to N objects.
+    """
+    tau = 0.70
+    delta = 1e-7
+    
+    # Run full adjusted pipeline on the dogs dataset
+    adjusted_kappa = calculate_adjusted_kappa_iou(T3_RATER_A, T3_RATER_B, tau=tau, delta=delta)
+    
+    # Verify that the nested geometric logging yields a stable bounded result
+    assert isinstance(adjusted_kappa, float)
+    assert adjusted_kappa > 0.0, f"Expected valid positive adjusted matrix score, got {adjusted_kappa}"