Merge pull request #85 from Paeti/enh/prepare_dataset_28

Enh/prepare dataset 28

.gitignore

@@ -102,3 +102,5 @@ venv.bak/
 
 # mypy
 .mypy_cache/
+mains/.idea
+preprocessing/.idea

mains/check_new_image.py

@@ -0,0 +1,57 @@
+import os, time
+import numpy as np
+import cv2 as cv
+
+face_cascade = cv.CascadeClassifier('haarcascade_frontalface_default.xml')
+delta = 5
+
+#change directories here
+path_to_watch = "tmp"
+output_dir = "Output/"
+
+before = dict ([(f, None) for f in os.listdir (path_to_watch)])
+while 1:
+    time.sleep (10)
+    print("Alive")
+    after = dict ([(f, None) for f in os.listdir (path_to_watch)])
+    added = [f for f in after if not f in before]
+    removed = [f for f in before if not f in after]
+    if added:
+        counter = 0
+        for element in added:
+            fp = path_to_watch + "/" + element
+            cur_pad = os.path.normpath(fp)
+            img = cv.imread(fp)
+            faces = face_cascade.detectMultiScale(img, 1.8, 5)
+            for (x, y, w, h) in faces:
+                counter += 1
+                roi_color = img[y:y + h, x:x + w]
+                cip = img[y - delta:y + h + delta, x - delta:x + w + delta].copy()
+
+                if counter > 0:
+                    facepath = output_dir + added[0] + str(counter) + ".jpg"
+                else:
+                    facepath = output_dir + added[0]
+                cur_path = os.path.normpath(facepath)
+                print("Created: " + cur_path)
+                try:
+
+                    cip = cv.resize(cip, (224, 224))
+                    b, g, r = cv.split(cip)
+                    b = b / 3
+                    g = g / 3
+                    r = r / 3
+                    meanRGB = cv.merge((b, g, r))
+                    cip = cip - meanRGB
+                    cv.imwrite(cur_path, cip)
+                    print("Written: " + cur_path)
+                except:
+                    print("Resizing failed! !ssize.empty()")
+            if os.path.exists(cur_pad):
+                os.remove(cur_pad)
+
+            else:
+                print("File not found")
+    if removed:
+        print("Removed: ", ", ".join (removed))
+    before = after

mains/extract_faces.py

@@ -0,0 +1,65 @@
+import numpy as np
+import cv2 as cv
+import os
+from skimage.io import imread_collection
+
+face_cascade = cv.CascadeClassifier('haarcascade_frontalface_default.xml')
+eye_cascade = cv.CascadeClassifier('haarcascade_eye.xml')
+
+delta = 5
+counter = 0
+
+#returns an array of all images in a given folder and its subfolders
+def load_images_from_folder(folder):
+    images = []
+    t = ()
+    for subfolder in os.listdir(folder):
+        test_img = cv.imread(os.path.join(folder, subfolder))
+        print(os.path.join(folder, subfolder))
+        if test_img is not None:
+            images.append(subfolder, test_img)
+        else:
+            for filename in os.listdir(os.path.join(folder, subfolder)):
+                print(os.path.join(folder, subfolder, filename))
+                imgs = cv.imread(os.path.join(folder, subfolder, filename))
+                if imgs is not None:
+                    images.append(filename, imgs)
+    return images
+
+# folder where to search for pictures - no non folder/picture files or error
+fp = "Inputfiles/imdb_crop"
+cur_pad = os.path.normpath(fp)
+image_tuples = load_images_from_folder(cur_pad)
+
+for img_tuple in image_tuples:
+    counter = 0
+    img_name = img_tuple[0]
+    img = img_tuple[1]
+
+    faces = face_cascade.detectMultiScale(img, 1.8, 5)
+    for (x,y,w,h) in faces:
+        counter += 1
+        roi_color = img[y:y+h, x:x+w]
+        cip = img[y-delta:y+h+delta, x-delta:x+w+delta].copy()
+
+        if counter > 0:
+            facepath = "Outputfiles/" + img_name + str(counter) + ".jpg"
+        else:
+            facepath = "Outputfiles/" + img_name
+        cur_path = os.path.normpath(facepath)
+        print(cur_path)
+        try:
+
+            cip = cv.resize(cip, (224,224))
+            b, g, r = cv.split(cip)
+            b = b / 3
+            g = g / 3
+            r = r / 3
+            meanRGB = cv.merge((b, g, r))
+            cip = cip - meanRGB
+            cv.imwrite(cur_path, cip)
+        except:
+            print("Resizing failed! !ssize.empty()")
+
+cv.waitKey(0)
+cv.destroyAllWindows()