Код: Выделить всё
from flask import Flask, render_template, request, jsonify
import os
import cv2
import numpy as np
from skimage.feature.texture import graycomatrix, graycoprops
from skimage.filters import gabor
from skimage.color import rgb2hsv
from skimage.segmentation import slic
from skimage.measure import regionprops
from sklearn.preprocessing import MinMaxScaler
import joblib
app = Flask(__name__)
# Directory to store uploaded files
UPLOAD_DIRECTORY = 'uploads'
RABIES_IMG_DIR = 'rabies_images'
NON_RABIES_IMG_DIR = 'non_rabies_images'
# Create the uploads directory if it doesn't exist
if not os.path.exists(UPLOAD_DIRECTORY):
os.makedirs(UPLOAD_DIRECTORY)
# Load trained model
clf_svm = joblib.load('SVM_model.joblib')
# Target size for image resizing
target_size = (224, 224)
# Function to preprocess an image
def preprocess_image(image_rgb, target_size):
try:
# Print type and shape of the input image
print("Input image - Type:", image_rgb.dtype, "Shape:", image_rgb.shape)
# Convert the image to grayscale
gray_image = cv2.cvtColor(image_rgb, cv2.COLOR_RGB2GRAY)
# Ensure the grayscale image is of type CV_8UC1
if gray_image.dtype != np.uint8:
gray_image = gray_image.astype(np.uint8)
# Print type and shape of the grayscale image
print("Grayscale image - Type:", gray_image.dtype, "Shape:", gray_image.shape)
# Contrast Enhancement (CLAHE)
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
enhanced_image = clahe.apply(gray_image)
# Convert the enhanced image back to RGB for further processing
enhanced_image_rgb = cv2.cvtColor(enhanced_image, cv2.COLOR_GRAY2RGB)
# Resize the enhanced image to the specified target size
img = cv2.resize(enhanced_image_rgb, target_size)
# Print type and shape of the preprocessed image
print("Preprocessed image - Type:", img.dtype, "Shape:", img.shape)
return img
except Exception as e:
print("Error during image preprocessing:", e)
return None
# Function to extract features from a preprocessed image
# Function to extract features from a preprocessed image
def extract_features(image_rgb, target_segments=100, max_iterations=20, max_features_range=(1000, 1100)):
try:
# Convert the RGB image to grayscale
gray_image = cv2.cvtColor(image_rgb, cv2.COLOR_RGB2GRAY)
# GLCM (Gray-Level Co-occurrence Matrix)
glcm = graycomatrix(gray_image, distances=[5], angles=[0], levels=256, symmetric=True, normed=True)
contrast = graycoprops(glcm, 'contrast')[0][0]
homogeneity = graycoprops(glcm, 'homogeneity')[0][0]
energy = graycoprops(glcm, 'energy')[0][0]
# Gabor Filters
gabor_features = []
for sigma in (1, 3):
for frequency in (0.05, 0.25):
theta = np.pi / 4. # Use a fixed value for theta
kernel = np.real(gabor(gray_image, frequency, theta=theta, sigma_x=sigma, sigma_y=sigma))
gabor_features.append(np.mean(kernel))
# Color Histograms (HSV color space)
hsv_image = rgb2hsv(image_rgb)
h_hist, _ = np.histogram(hsv_image[:, :, 0], bins=256, range=(0, 1))
s_hist, _ = np.histogram(hsv_image[:, :, 1], bins=256, range=(0, 1))
v_hist, _ = np.histogram(hsv_image[:, :, 2], bins=256, range=(0, 1))
# Color-Based Segmentation with adjusted parameters
for _ in range(max_iterations):
segments = slic(image_rgb, n_segments=target_segments, compactness=20, sigma=3, start_label=1)
num_segments = len(np.unique(segments))
if num_segments == target_segments:
break
elif num_segments < target_segments:
target_segments -= 1
else:
target_segments += 1
else:
print(f"Warning: Unable to generate {target_segments} segments after {max_iterations} attempts. Skipping prediction.")
return None
# Geometric Features
segment_props = regionprops(segments)
area_list = [prop.area for prop in segment_props]
perimeter_list = [prop.perimeter for prop in segment_props]
eccentricity_list = [prop.eccentricity for prop in segment_props]
# Debug prints to investigate the number of segments and the length of the area list
print(f"Number of segments: {num_segments}")
print(f"Length of area list: {len(area_list)}")
# Scale features to [0, 1] range
scaler = MinMaxScaler()
contrast_scaled = scaler.fit_transform([[contrast]])[0]
homogeneity_scaled = scaler.fit_transform([[homogeneity]])[0]
energy_scaled = scaler.fit_transform([[energy]])[0]
gabor_features_scaled = scaler.fit_transform([gabor_features])[0]
h_hist_scaled = scaler.fit_transform([h_hist])[0]
s_hist_scaled = scaler.fit_transform([s_hist])[0]
v_hist_scaled = scaler.fit_transform([v_hist])[0]
area_scaled = scaler.fit_transform([area_list])[0]
perimeter_scaled = scaler.fit_transform([perimeter_list])[0]
eccentricity_scaled = scaler.fit_transform([eccentricity_list])[0]
# Combine all features into a single feature vector
feature_vector = np.concatenate([contrast_scaled, homogeneity_scaled, energy_scaled,
gabor_features_scaled, h_hist_scaled, s_hist_scaled,
v_hist_scaled, area_scaled, perimeter_scaled,
eccentricity_scaled])
# Truncate or pad the feature vector to ensure it falls within the specified range
min_features, max_features = max_features_range
if len(feature_vector) > max_features:
feature_vector = feature_vector[:max_features]
elif len(feature_vector) < min_features:
feature_vector = np.pad(feature_vector, (0, min_features - len(feature_vector)), 'constant')
return np.array(feature_vector) # Convert to array
except Exception as e:
print(f"Error during feature extraction: {e}")
return None
# Function to classify the image as rabies or non-rabies
def classify_image(image_path):
# Read the uploaded image
user_image = cv2.imread(image_path)
# Preprocess the image
preprocessed_img = preprocess_image(user_image, target_size)
if preprocessed_img is not None:
# Extract features from preprocessed image
features = extract_features(preprocessed_img)
if features is not None:
# Make predictions using SVM model
svm_prediction = clf_svm.predict(features.reshape(1, -1))
if svm_prediction is not None:
return "Suspected of Rabies" if svm_prediction == 1 else "Not Suspected of Rabies"
else:
return "Prediction Error"
else:
return "Feature Extraction Error"
else:
return "Image Preprocessing Error"
# Route for home page
@app.route('/')
def index():
return render_template('index.html')
# Route for image upload and classification
@app.route('/upload', methods=['POST'])
def upload():
if 'file' not in request.files:
return jsonify({'error': 'No file part'})
file = request.files['file']
if file.filename == '':
return jsonify({'error': 'No selected file'})
# Save the uploaded file
file_path = os.path.join(UPLOAD_DIRECTORY, file.filename)
file.save(file_path)
# Classify the uploaded image
result = classify_image(file_path)
return jsonify({'result': result})
if __name__ == '__main__':
app.run(debug=True)
Я хочу, чтобы пользователь загрузил свое изображение собаки, и моя система будет различать его изображение с помощью двух имеющихся у меня каталогов, один из которых содержит изображения собак, больных бешенством. и другие, содержащие изображения собак, не связанных с бешенством. Таким образом, результатом будет то, есть ли у их собаки подозрение на бешенство или нет.
Подробнее здесь: https://stackoverflow.com/questions/783 ... ion-system
