Merge pull request #427 from the-silent-geek/Railway_Track_Fault_dete…

…ction

Track fault detection using DL

Railway Track Fault Detection/Model/conv.py

@@ -1,80 +1,80 @@
-# A Python script to change the size of images and save them inside a new directory
-# The script contains two functions:
-# 1. To  change the names of current images to standard number labels.
-# 2. To convert the images of random sizes to a fixed size 200X200 resolution.
-# This makes the further process of training the model easy as dimensions of each image
-# are the same
-
-
-# Import all the necessary modules
-import cv2
-import glob
-import os
-
-
-# Source folder location of images
-defective = "D://ML DL AI DSBDA//Railway Track Fault Detection//archive//images//Defective"
-nonDefective = "D://ML DL AI DSBDA//Railway Track Fault Detection//archive//images//Non Defective"
-# Create a separate directory to store processed images
-
-
-
-def changeNames(path: str) -> None:
-    'Change the naming conventions of images:'
-    k = 0
-    location = os.chdir(path)
-    for old_file_name in os.listdir(location):
-        new_file_name = '{}.jpg'.format(k)
-        os.rename(old_file_name, new_file_name)
-        k += 1
-
-
-def changeDimensionsForDefectiveImages(width: int, height: int) -> None:
-    'Change the random dimensions of all images to a fixed resolution and then insert into a new folder'
-    i = 0
-    for img in glob.glob(defective + '//*.jpg'):
-        image = cv2.imread(img)
-        imageResized = cv2.resize(image, (width, height))
-
-        # cv2.imshow('frame', image)
-
-        cv2.imwrite(
-            "D://ML DL AI DSBDA//Railway Track Fault Detection//Processed Images//Defective//000{}.jpg".format(i), imageResized)
-        cv2.waitKey(0)
-        i += 1
-
-        # if i == 2:
-        #     break
-
-    cv2.destroyAllWindows()
-
-def changeDimensionsForNonDefectiveImages(width: int, height: int) -> None:
-    'Change the random dimensions of all images to a fixed resolution and then insert into a new folder'
-    i = 0
-    for img in glob.glob(nonDefective + '//*.jpg'):
-        image = cv2.imread(img)
-        imageResized = cv2.resize(image, (width, height))
-
-        # cv2.imshow('frame', image)
-        cv2.imwrite(
-            "D://ML DL AI DSBDA//Railway Track Fault Detection//Processed Images//Non Defective//000{}.jpg".format(i), imageResized)
-        cv2.waitKey(0)
-        i += 1
-
-        # if i == 2:
-        #     break
-
-    cv2.destroyAllWindows()
-
-
-def main():
-    defectiveImagePath = 'D://ML DL AI DSBDA//Railway Track Fault Detection//archive//Images//Defective'
-    nonDefectivePath = 'D://ML DL AI DSBDA//Railway Track Fault Detection//archive//Images//Non Defective'
-    changeNames(defectiveImagePath)
-    changeNames(nonDefectivePath)
-    changeDimensionsForDefectiveImages(256, 256)
-    changeDimensionsForNonDefectiveImages(256, 256)
-
-
-if __name__ == '__main__':
-    main()
+# A Python script to change the size of images and save them inside a new directory
+# The script contains two functions:
+# 1. To  change the names of current images to standard number labels.
+# 2. To convert the images of random sizes to a fixed size 200X200 resolution.
+# This makes the further process of training the model easy as dimensions of each image
+# are the same
+
+
+# Import all the necessary modules
+import cv2
+import glob
+import os
+
+
+# Source folder location of images
+defective = "D://ML DL AI DSBDA//Railway Track Fault Detection//archive//images//Defective"
+nonDefective = "D://ML DL AI DSBDA//Railway Track Fault Detection//archive//images//Non Defective"
+# Create a separate directory to store processed images
+
+
+
+def changeNames(path: str) -> None:
+    'Change the naming conventions of images:'
+    k = 0
+    location = os.chdir(path)
+    for old_file_name in os.listdir(location):
+        new_file_name = '{}.jpg'.format(k)
+        os.rename(old_file_name, new_file_name)
+        k += 1
+
+
+def changeDimensionsForDefectiveImages(width: int, height: int) -> None:
+    'Change the random dimensions of all images to a fixed resolution and then insert into a new folder'
+    i = 0
+    for img in glob.glob(defective + '//*.jpg'):
+        image = cv2.imread(img)
+        imageResized = cv2.resize(image, (width, height))
+
+        # cv2.imshow('frame', image)
+
+        cv2.imwrite(
+            "D://ML DL AI DSBDA//Railway Track Fault Detection//Processed Images//Defective//000{}.jpg".format(i), imageResized)
+        cv2.waitKey(0)
+        i += 1
+
+        # if i == 2:
+        #     break
+
+    cv2.destroyAllWindows()
+
+def changeDimensionsForNonDefectiveImages(width: int, height: int) -> None:
+    'Change the random dimensions of all images to a fixed resolution and then insert into a new folder'
+    i = 0
+    for img in glob.glob(nonDefective + '//*.jpg'):
+        image = cv2.imread(img)
+        imageResized = cv2.resize(image, (width, height))
+
+        # cv2.imshow('frame', image)
+        cv2.imwrite(
+            "D://ML DL AI DSBDA//Railway Track Fault Detection//Processed Images//Non Defective//000{}.jpg".format(i), imageResized)
+        cv2.waitKey(0)
+        i += 1
+
+        # if i == 2:
+        #     break
+
+    cv2.destroyAllWindows()
+
+
+def main():
+    defectiveImagePath = 'D://ML DL AI DSBDA//Railway Track Fault Detection//archive//Images//Defective'
+    nonDefectivePath = 'D://ML DL AI DSBDA//Railway Track Fault Detection//archive//Images//Non Defective'
+    changeNames(defectiveImagePath)
+    changeNames(nonDefectivePath)
+    changeDimensionsForDefectiveImages(256, 256)
+    changeDimensionsForNonDefectiveImages(256, 256)
+
+
+if __name__ == '__main__':
+    main()

Track Fault Detection using DL/README.md

@@ -0,0 +1,53 @@
+# <h1 align = "center"> Track Fault Detection using DL </h1>
+## Aim of the project: 
+### The project focuses on detecting the faults in the railway tracks using various Deep Learning Algorithms
+### Dataset Link: [railway Track fault Detection Resized (224 X 224) | Kaggle](https://www.kaggle.com/datasets/gpiosenka/railway-track-fault-detection-resized-224-x-224)
+
+###  Libraries and Frameworks used:
+1. Pandas
+2. Numpy 
+3. Matplotlib
+4. Seaborn
+5. Tensorflow
+6. Keras
+7. visualkeras
+8. pickle
+9. pillow
+10. opendatasets
+
+
+## Deep Learning Algorithms used:
+1. Vgg16
+2. ResNet50
+3. InceptionV3
+4. DenseNet
+5. MobileNet
+
+## Accuracy and training time comparison of all the Deep Learning Algorithms
+|             |   Accuracy    |
+|-------------|---------------|
+|   Vgg16     |     96%       |
+|  ResNet50   |     71%       |
+| InceptionV3 |     94%       |  
+|  DenseNet   |     98%       |     
+|  MobileNet  |     97%       |
+
+# Accuracy plots of all models
+
+## Vgg16
+![vgg16](https://github.com/the-silent-geek/DL-Simplified/blob/55596e0bfc60b5aba38f5bb64519fa9363dd5d61/Railway%20Track%20Fault%20Detection/images/vgg16.png)
+
+## ResNet50
+![resnet](https://github.com/the-silent-geek/DL-Simplified/blob/55596e0bfc60b5aba38f5bb64519fa9363dd5d61/Railway%20Track%20Fault%20Detection/images/ResNet.png)
+
+## InceptionV3
+![inception](https://github.com/the-silent-geek/DL-Simplified/blob/55596e0bfc60b5aba38f5bb64519fa9363dd5d61/Railway%20Track%20Fault%20Detection/images/inceptionV3.png)
+
+## DenseNet
+![densenet](https://github.com/the-silent-geek/DL-Simplified/blob/55596e0bfc60b5aba38f5bb64519fa9363dd5d61/Railway%20Track%20Fault%20Detection/images/densenet.png)
+
+## MobileNet
+![mobilenet](https://github.com/the-silent-geek/DL-Simplified/blob/55596e0bfc60b5aba38f5bb64519fa9363dd5d61/Railway%20Track%20Fault%20Detection/images/mobileNet.png)
+
+# Conclusion
+DenseNet model performs better comparative to other models used on the above dataset.

Track Fault Detection using DL/dataset/README.md

@@ -0,0 +1,3 @@
+## You can download the dataset from the following Link 🔗 :
+
+### [https://www.kaggle.com/datasets/gpiosenka/railway-track-fault-detection-resized-224-x-224](url)