BACKEND :modified class_object_detection(method),added...

BACKEND :modified class_object_detection(method),added component_separation(method) & crop_and_image_resolution(method) in class_model_detection_service.py

backend/app.py

@@ -21,10 +21,14 @@ OUTPUTS_GENERATED_CLASS_FILES_PATH = os.path.join('outputs', 'generated_class_fi
 OUTPUTS_FOLDER = os.path.join(APP_ROOT, 'outputs')
 UML_GENERATOR_UPLOAD_FOLDER = os.path.join(APP_ROOT, 'uploads')
 SUBMISSION_PATH = os.path.join(APP_ROOT, 'submissions/use_case')
+SUBMISSION_PATH_CLASS = os.path.join(APP_ROOT, 'submissions/class')
+
 USE_CASE_SAVED_MODEL_PATH = os.path.join(APP_ROOT, 'tensorflow_models/use_case/model')
 USE_CASE_SAVED_LABEL_PATH = os.path.join(APP_ROOT, 'tensorflow_models/use_case/label')
 CLASS_SAVED_MODEL_PATH = os.path.join(APP_ROOT, 'tensorflow_models/class/model')
-CLASS_SAVED_LABEL_PATH = os.path.join(APP_ROOT, 'tensorflow_models/class/label')
+CLASS_SAVED_LABEL_PATH = os.path.join(APP_ROOT, 'tensorflow_models/class/label/label_map.pbtxt')
+CLASS_COMP_SAVED_MODEL_PATH = os.path.join(APP_ROOT, 'tensorflow_models/class_component/model')
+CLASS_COMP_SAVED_LABEL_PATH = os.path.join(APP_ROOT, 'tensorflow_models/class_component/label/label_map.pbtxt')
 app = Flask(__name__, static_folder='outputs')
 CORS(app)
 

backend/requirements.txt

 attrs==21.4.0

backend/services/class_model_detection_service.py

@@ -13,12 +13,11 @@ import tensorflow as tf
 
 from config.database import db
 
-def model_object_detection(filename, class_id):
-    detect_fn = tf.saved_model.load(app.CLASS_SAVED_MODEL_PATH)
-    category_index = label_map_util.create_category_index_from_labelmap(
-        app.CLASS_SAVED_LABEL_PATH + "/label_map.pbtxt",
-        use_display_name=True)
-    image_np = np.array(Image.open(app.SUBMISSION_PATH + '/' + filename))
+
+def class_object_detection(filename, class_comp_id, model_path, label_path, image_path):
+    detect_fn = tf.saved_model.load(model_path)
+    category_index = label_map_util.create_category_index_from_labelmap(label_path, use_display_name=True)
+    image_np = np.array(Image.open(image_path))
     input_tensor = tf.convert_to_tensor(image_np)
 
     input_tensor = input_tensor[tf.newaxis, ...]
@@ -32,8 +31,64 @@ def model_object_detection(filename, class_id):
 
     detections['detection_classes'] = detections['detection_classes'].astype(np.int64)
 
-    accurate_indexes = [k for k, v in enumerate(detections['detection_scores']) if (v > 0.4)]
+    accurate_indexes = [k for k, v in enumerate(detections['detection_scores']) if (v > 0.6)]
+    num_entities = len(accurate_indexes)
 
     class_id = operator.itemgetter(*accurate_indexes)(detections['detection_classes'])
     boxes = detections['detection_boxes']
-    text_extraction(filename, class_id, boxes, accurate_indexes, category_index, class_id)
\ No newline at end of file
+    return boxes, num_entities, accurate_indexes, num_entities, category_index, class_id
+
+
+def component_separation(filename, class_comp_id):
+    boxes, num_entities, accurate_indexes, num_entities, category_index, class_id = class_object_detection(filename,
+                                                                                                           class_comp_id,
+                                                                                                           app.CLASS_SAVED_MODEL_PATH,
+                                                                                                           app.CLASS_SAVED_LABEL_PATH,
+                                                                                                           app.SUBMISSION_PATH_CLASS + '/' + filename)
+    # Convert the class id in their name
+    if num_entities > 0:
+        for i in range(0, len(accurate_indexes)):
+
+            if category_index[class_id]['name'] == 'class':
+                class_image = crop_and_image_resolution(filename, boxes)
+                cv2.imwrite(app.SUBMISSION_PATH_CLASS, "class" + str(i))
+                # boxes,num_entities,accurate_indexes,num_entities,category_index,class_id = class_object_detection(class_image, class_comp_id, app.CLASS_SAVED_MODEL_PATH, app.CLASS_SAVED_LABEL_PATH,app.SUBMISSION_PATH_CLASS+'/'+"class"+str(i))
+
+            elif category_index[class_id]['name'] == 'interface':
+                image = crop_and_image_resolution(filename, boxes)
+
+
+def crop_and_image_resolution(filename, boxes):
+    image = cv2.imread(app.SUBMISSION_PATH + '/' + filename)
+    height, width, c = image.shape
+    # crop box format: xmin, ymin, xmax, ymax
+    ymin = boxes[0] * height
+    xmin = boxes[1] * width
+    ymax = boxes[2] * height
+    xmax = boxes[3] * width
+
+    cropped_image = image[int(ymin):int(ymax), int(xmin):int(xmax)]
+    image = cv2.imwrite('image.jpg', cropped_image)
+    # convert values to int
+    black = (0, 0, 0)
+    white = (255, 255, 255)
+    threshold = (160, 160, 160)
+
+    # Open input image in grayscale mode and get its pixels.
+    img = Image.open("image.jpg").convert("LA")
+    pixels = img.getdata()
+
+    new_pixels = []
+
+    # Compare each pixel
+    for pixel in pixels:
+        if pixel < threshold:
+            new_pixels.append(black)
+        else:
+            new_pixels.append(white)
+
+            # Create and save new image.
+    new_img = Image.new("RGB", img.size)
+    new_img.putdata(new_pixels)
+
+    return image

backend/tensorflow_models/class_component/label/label_map.pbtxt

+item {
+    id: 1
+    name: 'class_attributes'
+}
+
+item {
+    id: 2
+    name: 'class_methods'
+}
+
+item {
+    id: 3
+    name: 'class_name'
+}
\ No newline at end of file