This is the code of the model that I have implemented using direct CNN

CNN.txt

+# Import necessary libraries
+import matplotlib.pyplot as plt
+import numpy as np
+import cv2
+import os
+import PIL
+import tensorflow as tf
+from PIL import Image
+from tensorflow import keras
+from tensorflow.keras import layers
+from tensorflow.keras.models import Sequential
+
+# Define the path to the dataset directory
+data_dir='.\\datasets\\balls_images'
+
+# Print the path
+print(data_dir)
+
+# Convert path to a pathlib object
+import pathlib
+data_dir = pathlib.Path(data_dir)
+
+# Print the pathlib object
+print(data_dir)
+
+# List the image files in the dataset directory
+list(data_dir.glob('*/*.jpg'))[:5]
+
+# Count the total number of images in the dataset
+image_count = len(list(data_dir.glob('*/*.jpg')))
+print(image_count)
+
+# Define the list of legal ball images
+legal_ball = list(data_dir.glob('legal_ball/*'))
+
+# Display an example image from the legal ball images
+PIL.Image.open(str(legal_ball[1]))
+
+# Define a dictionary containing the paths of all ball images grouped by ball type
+balls_images_dict = {
+    'no_ball': list(data_dir.glob('no_ball/*')),
+    'legal_ball': list(data_dir.glob('legal_ball/*')),
+}
+
+# Define a dictionary containing the class labels for each ball type
+balls_labels_dict = {
+    'no_ball': 0,
+    'legal_ball': 1,
+}
+
+# Display the paths to the first five 'no_ball' images
+balls_images_dict['no_ball'][:5]
+
+# Load an image using OpenCV and display its shape
+img = cv2.imread(str(balls_images_dict['no_ball'][0]))
+img.shape
+
+# Resize the image and display its new shape
+cv2.resize(img,(180,180)).shape
+
+# Initialize empty lists for the features and labels
+X, y = [], []
+
+# Iterate through each ball type and its corresponding images
+for ball_type, images in balls_images_dict.items():
+    for image in images:
+        # Load the image using OpenCV and resize it to (224,224)
+        img = cv2.imread(str(image))
+        resized_img = cv2.resize(img,(224,224))
+        # Append the resized image to the feature list and its label to the label list
+        X.append(resized_img)
+        y.append(balls_labels_dict[ball_type])
+
+# Convert the feature and label lists to numpy arrays
+X = np.array(X)
+y = np.array(y)
+
+# Split the dataset into training and testing sets
+from sklearn.model_selection import train_test_split
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)
+
+# Normalize the pixel values of the training and testing sets
+X_train_scaled = X_train / 255
+X_test_scaled = X_test / 255
+
+# Define the number of classes
+num_classes = 2
+
+# Define the convolutional neural network model
+model = Sequential([
+  layers.Conv2D(16, 3, padding='same', activation='relu'),
+  layers.MaxPooling2D(),
+  layers.Conv2D(32, 3, padding='same', activation='relu'),
+  layers.MaxPooling2D(),
+  layers.Conv2D(64, 3, padding='same', activation='relu'),
+  layers.MaxPooling2D(),
+  layers.Flatten(),
+  layers.Dense(128, activation='relu'),
+  layers.Dense(num_classes)
+])
+
+# Compile the model
+model.compile(optimizer='adam',
+              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
+              metrics=['accuracy'])
+
+# Train the model
+model.fit(X_train_scaled, y_train, epochs=5)
+
+model.evaluate(X_test_scaled,y_test)
+
+loss, accuracy = model.evaluate(X_test_scaled,y_test)
+
+print('Accuracy:', accuracy)
+
+plt.imshow(X[0])
+
+# save the model
+model.save("balls_model.h5")
+
+
+from tensorflow.keras.models import load_model
+
+loaded_model = load_model("balls_model.h5")