From f719ff767aaa90048039586dfc7f161145304dd1 Mon Sep 17 00:00:00 2001
From: Benjamin Loison <12752145+Benjamin-Loison@users.noreply.github.com>
Date: Wed, 15 May 2024 20:15:20 +0200
Subject: [PATCH] #66: Stop execution if Pearson correlation is negative

---
 datasets/raise/attribute_source_camera.py              | 4 ++--
 datasets/raise/split_and_compare_prnus_of_subgroups.py | 4 ++--
 datasets/raise/utils.py                                | 8 ++++++--
 3 files changed, 10 insertions(+), 6 deletions(-)

diff --git a/datasets/raise/attribute_source_camera.py b/datasets/raise/attribute_source_camera.py
index e4efb97..23714fb 100755
--- a/datasets/raise/attribute_source_camera.py
+++ b/datasets/raise/attribute_source_camera.py
@@ -26,7 +26,7 @@ match DISTANCE:
         from rms_diff import rmsDiffNumpy
     case PEARSON_CORRELATION:
         import scipy
-        from utils import getPearsonCorrelation
+        from utils import getPearsonCorrelationDistance
 
 setting = ','.join([escapeFilePath(imageCameraFolder) for imageCameraFolder in IMAGES_CAMERAS_FOLDER]) + f'_{DENOISER}'
 
@@ -194,7 +194,7 @@ for computeExtremes in tqdm(([True] if minColor is None or maxColor is None else
                                 case ROOT_MEAN_SQUARE:
                                     distance = rmsDiffNumpy(cameraTestingImageNoise, camerasIterativeMean[camera].mean)
                                 case PEARSON_CORRELATION:
-                                    distance = getPearsonCorrelation(cameraTestingImagesNoise[actualCamera][cameraTestingImageIndex], camerasIterativeMean[camera].mean)
+                                    distance = getPearsonCorrelationDistance(cameraTestingImagesNoise[actualCamera][cameraTestingImageIndex], camerasIterativeMean[camera].mean)
                             print(f'{cameraTrainingImageIndex=} {cameraTestingImageIndex=} {camera=} {actualCamera=} {distance=}')
                             if minimalDistance is None or distance < minimalDistance:
                                 minimalDistance = distance
diff --git a/datasets/raise/split_and_compare_prnus_of_subgroups.py b/datasets/raise/split_and_compare_prnus_of_subgroups.py
index a42bdc5..b313193 100755
--- a/datasets/raise/split_and_compare_prnus_of_subgroups.py
+++ b/datasets/raise/split_and_compare_prnus_of_subgroups.py
@@ -20,7 +20,7 @@ match DISTANCE:
         from rms_diff import rmsDiffNumpy
     case PEARSON_CORRELATION:
         import scipy
-        from utils import getPearsonCorrelation
+        from utils import getPearsonCorrelationDistance
 
 setting = escapeFilePath(IMAGES_FOLDER) + f'_{DENOISER}'
 
@@ -67,7 +67,7 @@ for computeExtremes in tqdm(([True] if minColor is None or maxColor is None else
                     case ROOT_MEAN_SQUARE:
                         distance = rmsDiffNumpy(firstImage, secondImage)
                     case PEARSON_CORRELATION:
-                        distance = getPearsonCorrelation(firstImage, secondImage)
+                        distance = getPearsonCorrelationDistance(firstImage, secondImage)
                 distances += [distance]
 
 for subgroupIndex in range(NUMBER_OF_SUBGROUPS):
diff --git a/datasets/raise/utils.py b/datasets/raise/utils.py
index a3eb311..d92bf1b 100644
--- a/datasets/raise/utils.py
+++ b/datasets/raise/utils.py
@@ -43,8 +43,12 @@ class Distance(Enum)
     def getEndUserName(self):
         return self.name.replace('_', '').title()
 
-def getPearsonCorrelation(firstImage, secondImage):
-    return abs(scipy.stats.pearsonr(firstImage.flatten(), secondImage.flatten()).statistic - 1)
+def getPearsonCorrelationDistance(firstImage, secondImage):
+    pearsonCorrelation = scipy.stats.pearsonr(firstImage.flatten(), secondImage.flatten()).statistic
+    if pearsonCorrelation < 0:
+        print('Negative Pearson correlation, investigate further!')
+        exit(1)
+    return abs(pearsonCorrelation - 1)
 
 # Source: https://stackoverflow.com/a/43346070
 def getGaussianKernel(size, sigma):