adding extract_information single image test and segmentation file

extract_information.py

@@ -0,0 +1,63 @@
+import cv2
+import easyocr
+import numpy as np
+
+def extract_information_from_image(img, coordinates_dict, **kwargs):
+
+    def preprocess_image(image):
+        image = cv2.medianBlur(image, 3)
+        image = image * 0.7
+        image = np.clip(image, 0, 255)
+        image = image.astype(np.uint8)
+        alpha = 1.5
+        beta = 0
+        image = np.clip(alpha * image + beta, 0, 255).astype(np.uint8)
+        image = cv2.medianBlur(image, 5)
+        image = cv2.GaussianBlur(image, (3, 3), 0)
+        return image
+
+
+    allowed_list = " 0123456789(){}[]\'\"\\/.,_"
+    def check(input:str):
+        for letter in input:
+            if not (letter in allowed_list):
+    #             print('check list: ', letter)
+                return '--'
+        return input
+
+
+    if kwargs.get("reader") is None:
+        reader = easyocr.Reader(["en"])
+    else:
+        reader = kwargs.get("reader")
+
+    img = preprocess_image(img)
+    img = cv2.resize(img, (1280, 720))
+    #cv2.imwrite("saved_test_3.jpg", img=img)
+
+    #print(coordinates_dict)
+
+    res = {}
+    for key, value in coordinates_dict.items():
+        temp = reader.readtext(img[value[1]:value[3], value[0]:value[2]], detail=0)
+        print(temp)
+        try:
+            res[key] = (
+                check(input=temp[0])
+                if len(temp) >=1 else '--'
+            )
+        except:
+            res[key] = '--'
+    return res
+
+if __name__ == "__main__":
+    img = cv2.imread("ex.jpeg")
+    print(extract_information_from_image(img, {
+    "map": [141, 378, 272, 447],
+    "heart_rate": [71, 80, 415, 242],
+    "spo2": [93, 469, 424, 605],
+    "sbp": [18, 310, 195, 405],
+    "rr": [303, 646, 452, 726],
+    "dbp": [257, 317, 442, 398]
+  }))
+

pipeline-1-single-image.py

@@ -0,0 +1,45 @@
+from ultralytics import YOLO
+from classifier_model import predict_monitor
+from segmentation2polygon import segmentation2polygon, do_perspective_transformation
+import cv2
+from classifier_model import predict_monitor
+import json
+from extract_information import extract_information_from_image
+import argparse
+
+with open('data.json', 'r') as f:
+	content = json.load(f)
+parser = argparse.ArgumentParser(description="CloudPhysician")
+parser.add_argument('-i', '--img', help='Path to image')
+
+def extract_info(image):
+	yolo_model = YOLO("./model_yolo/best.pt")
+	yolo_result = yolo_model.predict(image)
+
+	yolo_mask = yolo_result[0].masks.xy
+
+	approx_polygon = segmentation2polygon(yolo_mask[0]).reshape((-1, 2))
+
+	perspective_image = do_perspective_transformation(image, approx_polygon)
+
+	class_number = predict_monitor(perspective_image)
+	if class_number == "1":
+		coordinates_dict = content['first']
+	elif class_number =="2":
+		coordinates_dict = content['second']
+	elif class_number =="3":
+		coordinates_dict = content['third']
+	elif class_number =="4":
+		coordinates_dict = content['fourth']
+	else:
+		exit(0)
+	return extract_information_from_image(image, coordinates_dict)
+
+
+
+if __name__ == '__main__':
+	args=parser.parse_args()
+	img = cv2.imread(args.img)
+	print(extract_info(img))
+
+

segmentation2polygon.py

@@ -77,7 +77,7 @@ def pipeline():
     segmentation = get_segmentation()
     segmentation = convert_str_to_array(segmentation)
     img_size = get_height_and_width_of_img(
-        "./yolo/yolo_predictions/medicakolkata_rcu_mon--8_2022_5_18_17_5_38.jpeg"
+        "ex.jpeg"
     )
     polygon = from_normalised_to_pixel(img_size, segmentation)
     polygon = segmentation2polygon(polygon)
@@ -92,7 +92,7 @@ def draw_points(polygon) -> None:
     Output: None
     """
     img = cv2.imread(
-        "./yolo/yolo_predictions/medicakolkata_rcu_mon--8_2022_5_18_17_5_38.jpeg"
+        "ex.jpeg"
     )
     for i in polygon:
         cv2.circle(img, (i[0], i[1]), 5, (0, 0, 255), -1)
@@ -129,15 +129,22 @@ def do_perspective_transformation(image, input_array):
 
 
 if __name__ == "__main__":
-    approx_polygon = pipeline().reshape((-1, 2))
+    '''approx_polygon = np.array([[  42 , 12],
+ [  72 , 692],
+ [1252 , 684],
+ [1236 ,  10]])
+
     print(len(approx_polygon))
     draw_points(approx_polygon)
     perspective = do_perspective_transformation(
         cv2.imread(
-            "./yolo/yolo_predictions/medicakolkata_rcu_mon--8_2022_5_18_17_5_38.jpeg"
+            "ex.jpeg"
         ),
         approx_polygon,
     )
     cv2.imshow("perspective", perspective)
+    #cv2.imwrite('ex.jpeg', perspective)
     cv2.waitKey(0)
-    cv2.destroyAllWindows()
+    cv2.destroyAllWindows()'''
+
+    print(segmentation2polygon(get_segmentation()))

yolo.py

@@ -0,0 +1,21 @@
+"""
+ TODO: 
+    1. get the .pt file of the yolo model and try to make it in such a way that it store output in our desired format
+"""
+from ultralytics import YOLO
+from segmentation2polygon import segmentation2polygon, draw_points
+
+model = YOLO("./model_yolo/best.pt")
+results = model.predict("ex.jpeg")
+
+mask = results[0].masks
+
+mask = mask.xy
+print(mask[0])
+segmentation = segmentation2polygon(mask[0])
+print(segmentation.reshape((-1, 2)))
+#with open('mask.txt', 'w') as f:
+    #f.write(mask[0])
+draw_points(segmentation.reshape((-1, 2)))
+
+