#62: Add benchmark_load_part_of_images.py

faster

datasets/raise/attribute_source_camera.py

@@ -7,7 +7,6 @@ from utils import denoise, iterativeMean, getColorChannel, escapeFilePath, Color
 import sys
 import os
 import random
-import scipy
 
 sys.path.insert(0, '../../algorithms/distance/')
 
@@ -44,13 +43,14 @@ camerasIterativeMean = {camera: iterativeMean() for camera in IMAGES_CAMERAS_FOL
 # Compute the minimal color channel camera resolution.
 # Assume that for each camera, its images have the same resolution.
 # The following consider a given color channel resolution, assuming they all have the same resolution.
-minimalColorChannelCameraResolution = None
-for camera in IMAGES_CAMERAS_FOLDER:
-    imageFileName = imagesCamerasFileNames[camera][0]
-    imageFilePath = f'{IMAGES_CAMERAS_FOLDER[camera]}/{imageFileName}'
-    singleColorChannelImagesShape = getColorChannel(imageFilePath, Color.RED).shape
-    if minimalColorChannelCameraResolution is None or singleColorChannelImagesShape < minimalColorChannelCameraResolution:
-        minimalColorChannelCameraResolution = singleColorChannelImagesShape
+minimalColorChannelCameraResolution = (100, 100)#None
+if minimalColorChannelCameraResolution is None:
+    for camera in IMAGES_CAMERAS_FOLDER:
+        imageFileName = imagesCamerasFileNames[camera][0]
+        imageFilePath = f'{IMAGES_CAMERAS_FOLDER[camera]}/{imageFileName}'
+        singleColorChannelImagesShape = getColorChannel(imageFilePath, Color.RED).shape
+        if minimalColorChannelCameraResolution is None or singleColorChannelImagesShape < minimalColorChannelCameraResolution:
+            minimalColorChannelCameraResolution = singleColorChannelImagesShape
 
 minColor = 0#None
 maxColor = 7952#None
@@ -75,7 +75,18 @@ def getMultipleColorsImage(singleColorChannelImages):
     return multipleColorsImage
 
 def getImagePrnuEstimateNpArray(singleColorChannelImages, multipleColorsImage, camera):
-    singleColorChannelDenoisedImages = {color: denoise(singleColorChannelImages[color], DENOISER) if DENOISER != Denoiser.MEAN else cameraColorMeans[camera][color] for color in Color}
+    singleColorChannelDenoisedImages = {}
+    for color in Color:
+        if DENOISER != Denoiser.MEAN:
+            singleColorChannelDenoisedImage = denoise(singleColorChannelImages[color], DENOISER)
+        else:
+            cameraColorMean = cameraColorMeans[camera][color]
+            if PREDICT_ONLY_ON_WHOLE_TRAINING_SET:
+                singleColorChannelDenoisedImage = cameraColorMean
+            else:
+                cameraColorCurrentMean = cameraColorMean.mean
+                singleColorChannelDenoisedImage = cameraColorCurrentMean
+        singleColorChannelDenoisedImages[color] = singleColorChannelDenoisedImage
     multipleColorsDenoisedImage = mergeSingleColorChannelImagesAccordingToBayerFilter(singleColorChannelDenoisedImages)
     imagePrnuEstimateNpArray = multipleColorsImage - multipleColorsDenoisedImage
     return imagePrnuEstimateNpArray
@@ -85,13 +96,24 @@ imagesCamerasFilePaths = {camera: [f'{IMAGES_CAMERAS_FOLDER[camera]}/{imagesCame
 # If `PREDICT_ONLY_ON_WHOLE_TRAINING_SET`, then compute the means of camera images to empower the `MEAN` denoiser.
 # Otherwise initialize these means of camera images to `iterativeMean`.
 if DENOISER == Denoiser.MEAN:
-    cameraColorMeans = {camera: (getColorMeans(imagesCamerasFilePaths[camera][:numberOfTrainingImages], Color, minimalColorChannelCameraResolution) if PREDICT_ONLY_ON_WHOLE_TRAINING_SET else iterativeMean()) for camera in imagesCamerasFilePaths}
+    cameraColorMeans = {camera: (getColorMeans(imagesCamerasFilePaths[camera][:numberOfTrainingImages], Color, minimalColorChannelCameraResolution) if PREDICT_ONLY_ON_WHOLE_TRAINING_SET else {color: iterativeMean() for color in Color}) for camera in imagesCamerasFilePaths}
 
 from utils import silentTqdm
 #tqdm = silentTqdm
 
 returnSingleColorChannelImage = lambda singleColorChannelImage, _minColor, _maxColor: singleColorChannelImage
 
+# Assume to have `{min,max}Color` hardcoded.
+print('Load training images to memory')
+rescaleIfNeeded = rescaleRawImageForDenoiser
+cameraTrainingImages = {}
+for cameraTrainingImageIndex in tqdm(range(numberOfTrainingImages), 'Camera training image index'):
+    for cameraIndex, camera in enumerate(tqdm(IMAGES_CAMERAS_FOLDER, 'Camera')):
+        singleColorChannelImages = getSingleColorChannelImages(camera, cameraTrainingImageIndex)
+        multipleColorsImage = getMultipleColorsImage(singleColorChannelImages)
+        cameraTrainingImages[camera] = cameraTrainingImages.get(camera, []) + [multipleColorsImage]
+print('Training images loaded to memory')
+
 # 2 loops:
 # - the first one is about computing `{min,max}Color`
 # - the second one is about estimating better and better the PRNU of each camera, as consider more and more training images and measuring the resulting attribution of cameras
@@ -123,10 +145,17 @@ for computeExtremes in tqdm(([True] if minColor is None or maxColor is None else
                 minColor, maxColor = updateExtremes(multipleColorsImage, minColor, maxColor)
                 continue
 
+            if DENOISER == Denoiser.MEAN:
+                for color in Color:
+                    cameraColorMeans[camera][color].add(singleColorChannelImages[color])
             imagePrnuEstimateNpArray = getImagePrnuEstimateNpArray(singleColorChannelImages, multipleColorsImage, camera)
 
             cameraIterativeMean = camerasIterativeMean[camera]
-            cameraIterativeMean.add(imagePrnuEstimateNpArray)
+            if DENOISER != Denoiser.MEAN or PREDICT_ONLY_ON_WHOLE_TRAINING_SET:
+                cameraIterativeMean.add(imagePrnuEstimateNpArray)
+            else:
+                # Still use `cameraIterativeMean` to simplify the implementation.
+                cameraIterativeMean.mean = np.mean([cameraTrainingImage - mergeSingleColorChannelImagesAccordingToBayerFilter({color: cameraColorMeans[camera][color].mean for color in Color}) for cameraTrainingImage in cameraTrainingImages[camera][:cameraTrainingImageIndex + 1]], axis = 0)
             # If we are considering the last camera and (not `PREDICT_ONLY_ON_WHOLE_TRAINING_SET` or we are considering the last training image), then we proceeded an additional image for all cameras and we can predict the accuracy at this learning step.
             if cameraIndex == numberOfCameras - 1 and (not PREDICT_ONLY_ON_WHOLE_TRAINING_SET or cameraTrainingImageIndex == numberOfTrainingImages - 1):
                 numberOfTrainingImagesAccuracy = 0
@@ -139,8 +168,15 @@ for computeExtremes in tqdm(([True] if minColor is None or maxColor is None else
                         #plt.imsave(f'{escapeFilePath(actualCamera)}_{cameraTestingImageIndex}.png', cameraTestingImagesNoise[actualCamera][cameraTestingImageIndex])
                         # Loop over each camera to compute closeness between the considered testing image noise and the estimated PRNUs of the various cameras.
                         for camera in IMAGES_CAMERAS_FOLDER:
-                            distance = abs(scipy.stats.pearsonr(cameraTestingImagesNoise[actualCamera][cameraTestingImageIndex].flatten(), camerasIterativeMean[camera].mean.flatten()).statistic - 1)
-                            print(f'{cameraTestingImageIndex=} {camera=} {actualCamera=} {distance=}')
+                            if DENOISER != Denoiser.MEAN:
+                                cameraTestingImageNoise = cameraTestingImagesNoise[actualCamera][cameraTestingImageIndex]
+                            else:
+                                singleColorChannelImages = getSingleColorChannelImages(camera, numberOfTrainingImages + cameraTestingImageIndex)
+                                multipleColorsImage = getMultipleColorsImage(singleColorChannelImages)
+                                cameraTestingImageNoise = getImagePrnuEstimateNpArray(singleColorChannelImages, multipleColorsImage, camera)
+
+                            distance = rmsDiffNumpy(cameraTestingImageNoise, camerasIterativeMean[camera].mean)
+                            print(f'{cameraTrainingImageIndex=} {cameraTestingImageIndex=} {camera=} {actualCamera=} {distance=}')
                             if minimalDistance is None or distance < minimalDistance:
                                 minimalDistance = distance
                                 cameraPredicted = camera

datasets/raise/benchmark_load_part_of_images.py

@@ -0,0 +1,41 @@
+#!/usr/bin/env python
+
+from utils import getColorChannel, Color
+import os
+from tqdm import tqdm
+import numpy as np
+from enum import Enum, auto
+
+IMAGES_CAMERAS_FOLDER = {
+    'RAISE': 'flat-field/nef',
+    'Rafael 23/04/24': 'rafael/230424',
+}
+
+class Operation(Enum):
+    LOAD_RAW = auto()
+    SAVE = auto()
+    LOAD_NPY = auto()
+
+OPERATION = Operation.LOAD_RAW
+RESOLUTION = 100
+
+for camera in tqdm(IMAGES_CAMERAS_FOLDER, 'Camera'):
+    imagesCameraFolder = IMAGES_CAMERAS_FOLDER[camera]
+    for file in tqdm(os.listdir(imagesCameraFolder), 'Image'):
+        if file.endswith('.NEF'):
+            #print(file)
+            imageFilePath = f'{imagesCameraFolder}/{file}'
+            numpyFilePath = f'{imageFilePath}.npy'
+            for color in Color:
+                if OPERATION in [Operation.LOAD_RAW, Operation.SAVE]:
+                    rawColorChannel = getColorChannel(imageFilePath, color)
+                if OPERATION == Operation.SAVE:
+                    np.save(numpyFilePath, {color: rawColorChannel})
+                if OPERATION == Operation.LOAD_NPY:
+                    rawColorChannel = np.load(numpyFilePath, allow_pickle = True)
+                    #print(type(rawColorChannel))
+                    #print(rawColorChannel)
+                    #print(rawColorChannel.shape)
+                    print(rawColorChannel.item()[color].mean())
+                break
+    break