add app.py
Browse files
app.py
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Libraries
|
2 |
+
import streamlit as st
|
3 |
+
from tensorflow import keras
|
4 |
+
import tensorflow as tf
|
5 |
+
from PIL import Image
|
6 |
+
import numpy as np
|
7 |
+
import cv2
|
8 |
+
import matplotlib.pyplot as plt
|
9 |
+
from imutils import perspective
|
10 |
+
from scipy.spatial import distance as dist
|
11 |
+
import os
|
12 |
+
|
13 |
+
# Constants
|
14 |
+
MODEL_PATH = 'model.h5'
|
15 |
+
IMAGE_DIR = 'images'
|
16 |
+
|
17 |
+
# Functions
|
18 |
+
def load_image(image_file):
|
19 |
+
img = Image.open(image_file)
|
20 |
+
return img
|
21 |
+
|
22 |
+
def midpoint(ptA, ptB):
|
23 |
+
return ((ptA[0] + ptB[0]) /2 , (ptA[1] + ptB[1]) /2)
|
24 |
+
|
25 |
+
def draw_dimensions(orig_image,predict_image,erode_iteration,open_iteration):
|
26 |
+
kernel1 =( np.ones((5,5), dtype=np.float32))#
|
27 |
+
kernel_sharpening = np.array([[-1,-1,-1],
|
28 |
+
[-1,9,-1],
|
29 |
+
[-1,-1,-1]])#
|
30 |
+
|
31 |
+
image = predict_image
|
32 |
+
image2 = orig_image
|
33 |
+
|
34 |
+
image = cv2.morphologyEx(image, cv2.MORPH_OPEN, kernel1,iterations=open_iteration )
|
35 |
+
image = cv2.filter2D(image, -1, kernel_sharpening)
|
36 |
+
image = cv2.erode(image,kernel1,iterations =erode_iteration)
|
37 |
+
|
38 |
+
image=cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)#original
|
39 |
+
|
40 |
+
thresh = cv2.threshold(image, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)[1]
|
41 |
+
labels=cv2.connectedComponents(thresh,connectivity=8)[1]
|
42 |
+
a=np.unique(labels)
|
43 |
+
count2=0
|
44 |
+
for label in a:
|
45 |
+
if label == 0:
|
46 |
+
continue
|
47 |
+
|
48 |
+
# Create a mask
|
49 |
+
mask = np.zeros(thresh.shape, dtype="uint8")
|
50 |
+
mask[labels == label] = 255
|
51 |
+
# Find contours and determine contour area
|
52 |
+
cnts,hieararch = cv2.findContours(mask.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
|
53 |
+
cnts = cnts[0]
|
54 |
+
c_area = cv2.contourArea(cnts)
|
55 |
+
|
56 |
+
(x,y),radius = cv2.minEnclosingCircle(cnts)
|
57 |
+
rect = cv2.minAreaRect(cnts)
|
58 |
+
box = cv2.boxPoints(rect)
|
59 |
+
box = np.array(box, dtype="int")
|
60 |
+
box = perspective.order_points(box)
|
61 |
+
color1 = (list(np.random.choice(range(150), size=3)))
|
62 |
+
color =[int(color1[0]), int(color1[1]), int(color1[2])]
|
63 |
+
cv2.drawContours(image2,[box.astype("int")],0,color,2)
|
64 |
+
(tl,tr,br,bl)=box
|
65 |
+
|
66 |
+
(tltrX,tltrY)=midpoint(tl,tr)
|
67 |
+
(blbrX,blbrY)=midpoint(bl,br)
|
68 |
+
# compute the midpoint between the top-left and top-right points,
|
69 |
+
# followed by the midpoint between the top-righ and bottom-right
|
70 |
+
(tlblX,tlblY)=midpoint(tl,bl)
|
71 |
+
(trbrX,trbrY)=midpoint(tr,br)
|
72 |
+
# draw the midpoints on the image
|
73 |
+
cv2.circle(image2, (int(tltrX), int(tltrY)), 5, (255, 0, 0), -1)
|
74 |
+
cv2.circle(image2, (int(blbrX), int(blbrY)), 5, (255, 0, 0), -1)
|
75 |
+
cv2.circle(image2, (int(tlblX), int(tlblY)), 5, (255, 0, 0), -1)
|
76 |
+
cv2.circle(image2, (int(trbrX), int(trbrY)), 5, (255, 0, 0), -1)
|
77 |
+
cv2.line(image2, (int(tltrX), int(tltrY)), (int(blbrX), int(blbrY)),color, 2)
|
78 |
+
cv2.line(image2, (int(tlblX), int(tlblY)), (int(trbrX), int(trbrY)),color, 2)
|
79 |
+
dA = dist.euclidean((tltrX, tltrY), (blbrX, blbrY))
|
80 |
+
dB = dist.euclidean((tlblX, tlblY), (trbrX, trbrY))
|
81 |
+
|
82 |
+
global dimA
|
83 |
+
dimA = dA*0.08
|
84 |
+
global dimB
|
85 |
+
dimB = dB*0.08
|
86 |
+
cv2.putText(image2, "{:.1f} millimeter".format(dimA),(int(tltrX - 10), int(tltrY - 10)), cv2.FONT_HERSHEY_SIMPLEX,0.65, (0,0,0), 2)
|
87 |
+
cv2.putText(image2, "{:.1f} millimeter".format(dimB),(int(trbrX + 10), int(trbrY+10)), cv2.FONT_HERSHEY_SIMPLEX,0.65,(0,0,0), 2)
|
88 |
+
|
89 |
+
return image2
|
90 |
+
|
91 |
+
def segment_molar(image_file):
|
92 |
+
|
93 |
+
img=load_image(image_file)
|
94 |
+
|
95 |
+
img = np.asarray(img)
|
96 |
+
img = cv2.resize(img, (512, 256))
|
97 |
+
img = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
|
98 |
+
img = np.expand_dims(img, axis=0)
|
99 |
+
|
100 |
+
prediction = model.predict(img)
|
101 |
+
|
102 |
+
output = prediction.reshape(256,512)
|
103 |
+
|
104 |
+
return img , output
|
105 |
+
|
106 |
+
def measure_molar(image_file):
|
107 |
+
|
108 |
+
img=load_image(image_file)
|
109 |
+
|
110 |
+
img = np.asarray(img)
|
111 |
+
img_prd= cv2.resize(img , (512, 256))
|
112 |
+
img_prd = cv2.cvtColor(img_prd, cv2.COLOR_RGB2GRAY)
|
113 |
+
img_prd = np.expand_dims(img_prd, axis=0)
|
114 |
+
prediction = model.predict(img_prd)
|
115 |
+
prediction=prediction*255
|
116 |
+
prediction = prediction.astype("uint8")
|
117 |
+
prediction_img=prediction.reshape(256,512)
|
118 |
+
img2 = np.zeros( ( np.array(prediction_img).shape[0], np.array(prediction_img).shape[1], 3 ) )
|
119 |
+
img2[:,:,0] = prediction_img
|
120 |
+
img2[:,:,1] = prediction_img
|
121 |
+
img2[:,:,2] = prediction_img
|
122 |
+
img2 = img2.astype("uint8")
|
123 |
+
predicted = cv2.resize(img2 , (img.shape[1],img.shape[0]), interpolation=cv2.INTER_LANCZOS4)
|
124 |
+
output=draw_dimensions(img,predicted,3,2)
|
125 |
+
|
126 |
+
return output
|
127 |
+
|
128 |
+
# Load model and images
|
129 |
+
model = keras.models.load_model(MODEL_PATH)
|
130 |
+
images = os.listdir(IMAGE_DIR)
|
131 |
+
images = [f'{IMAGE_DIR}/image' for image in images]
|
132 |
+
|
133 |
+
# UserInterface
|
134 |
+
st.header("3rdMolar Segmentation")
|
135 |
+
st.subheader("Select Image:")
|
136 |
+
image_file = st.selectbox('Select Image' , images)
|
137 |
+
|
138 |
+
if image_file is not None:
|
139 |
+
|
140 |
+
img , output1 = segment_molar(image_file)
|
141 |
+
output2 = measure_molar(image_file)
|
142 |
+
|
143 |
+
st.image(img,width=850)
|
144 |
+
st.image(output1,width=850)
|
145 |
+
st.image(output2,width=850)
|
146 |
+
|