Upload 5 files

MLathon.py this file contains the core architecture and functionality of the model. Run this in appropriate epochs to fit the model.
improve.py this file contains code to load a saved model to finr-tune it further. Use effective optimizers to improve the learning process.
test.py use this file to test the model in an matplotlib figure
eval.py use this file to test it with testing dataset.

MLathon.py

@@ -0,0 +1,94 @@
+import tensorflow as tf
+import os
+
+import pathlib
+import numpy as np
+data_dir = pathlib.Path("/Users/rosh/Downloads/Train_data")
+class_names = np.array(sorted([item.name for item in data_dir.glob("*")]))
+class_names = list(class_names)
+import matplotlib.pyplot as plt
+import matplotlib.image as mpimg
+import random
+def view_random_image(target_dir, target_class):
+
+  target_folder = target_dir + "/" + target_class
+
+  random_image = random.sample(os.listdir(target_folder), 1)
+
+  img = mpimg.imread(target_folder + "/" + random_image[0])
+  plt.imshow(img)
+  plt.title(target_class)
+  plt.axis("off")
+
+  print(f"Image shape: {img.shape}")
+  plt.show()
+  return img
+
+
+#img = view_random_image(target_dir="/Users/rosh/Downloads/Train_data",target_class)
+
+from tensorflow.keras.preprocessing.image import ImageDataGenerator
+
+tf.random.set_seed(42)
+
+
+# Define data augmentation parameters
+train_datagen = ImageDataGenerator(
+    rotation_range=20,      # Random rotation in the range [-20, 20] degrees
+    width_shift_range=0.1,  # Random horizontal shift by up to 10% of the width
+    height_shift_range=0.1, # Random vertical shift by up to 10% of the height
+    shear_range=0.2,        # Shear intensity (shear angle in radians)
+    zoom_range=0.2,         # Random zoom in the range [0.8, 1.2]
+    horizontal_flip=True,   # Random horizontal flipping
+    vertical_flip=True,     # Random vertical flipping
+    fill_mode='nearest',    # Fill mode for points outside the input boundaries
+    rescale=1./255          # Rescaling factor
+)
+
+valid_datagen = ImageDataGenerator(
+    rescale=1./255          # Rescaling factor
+)
+
+
+train_dir = "/Users/rosh/Downloads/Train_data"
+valid_dir = "/Users/rosh/Downloads/Validation_data"
+
+train_data = train_datagen.flow_from_directory(directory=train_dir,
+                                               batch_size=32,
+                                               target_size=(224, 224),
+                                               class_mode="categorical",
+                                               seed=42)
+valid_data = valid_datagen.flow_from_directory(directory=valid_dir,
+                                               batch_size=32,
+                                               target_size=(224, 224),
+                                               class_mode="categorical",
+                                               seed=42)
+
+
+model_1 = tf.keras.Sequential([
+    tf.keras.layers.Conv2D(filters=32, kernel_size=3, activation="relu", input_shape=(224, 224, 3)),
+    tf.keras.layers.MaxPool2D(pool_size=2, padding="valid"),
+    tf.keras.layers.Conv2D(64, 3, activation="relu"),
+    tf.keras.layers.MaxPool2D(2),
+    tf.keras.layers.Conv2D(64, 3, activation="relu"),
+    tf.keras.layers.MaxPool2D(2),
+    tf.keras.layers.Flatten(),
+    tf.keras.layers.Dense(64, activation="relu"),
+    tf.keras.layers.Dense(10, activation="softmax")
+])
+
+
+model_1.compile(loss=tf.keras.losses.CategoricalCrossentropy(),
+                optimizer=tf.keras.optimizers.Adam(),
+                metrics=["accuracy"])
+
+
+history = model_1.fit(train_data,
+                      epochs=100,
+                      steps_per_epoch=len(train_data),
+                      validation_data=valid_data,
+                      validation_steps=len(valid_data),
+                      verbose=1)
+
+model_1.save("model_3.h5")
+#

count.py

@@ -0,0 +1,34 @@
+# from tensorflow.keras.preprocessing.image import ImageDataGenerator
+# import numpy as np
+# import tensorflow as tf
+# valid_datagen = ImageDataGenerator(
+#     rescale=1./255          # Rescaling factor
+# )
+# valid_dir = "/Users/rosh/Downloads/Validation_data"
+# valid_data = valid_datagen.flow_from_directory(directory=valid_dir,
+#                                                batch_size=32,
+#                                                target_size=(224, 224),
+#                                                class_mode="categorical",
+#                                                seed=42)
+# loaded_model = tf.keras.models.load_model('improved_model_4.h5')
+# true_labels = []
+# for i in range(len(valid_data)):
+#     _, labels = valid_data[i]
+#     true_labels.extend(np.argmax(labels, axis=1))
+#
+# # Print true labels
+# print("True labels:", true_labels)
+# pred_prob = loaded_model.predict(valid_data)
+# preds = pred_prob.argmax(axis=1)
+# print("Predicted: ")
+# count = 0
+# for i in range(len(preds)):
+#     if true_labels[i] == preds[i]:
+#         count += 1
+# print(count)
+#print(tf.keras.models.load_model('model_4_improved_1.h5').summary())
+import keras
+import tensorflow as tf
+
+print("Keras version:", keras.__version__)
+print("TensorFlow version:", tf.__version__)

eval.py

@@ -0,0 +1,38 @@
+import os
+import csv
+import numpy as np
+from tensorflow.keras.preprocessing import image
+import tensorflow as tf
+# Define the path to the directory containing test images
+predict_dir = "/Users/rosh/Downloads/Eval_data"  # Change this to the actual path
+model = tf.keras.models.load_model("model_4_improved_8.h5")
+# Define class labels
+class_labels = ['Crane', 'Crow', 'Egret', 'Kingfisher','Myna','Peacock','Pitta','Rosefinch','Tailorbird','Wagtail']
+# Open a CSV file to write the results
+with open('pred.csv', mode='w', newline='') as csvfile:
+    writer = csv.writer(csvfile)
+    writer.writerow(['Name', 'Target_name','Target_num'])
+    qq=0
+    # Loop through each image file and make predictions
+    for img_file in os.listdir(predict_dir):
+        print(qq)
+        # Load and preprocess the image
+        img_path = '/Users/rosh/Downloads/Eval_data'+'/'+ img_file
+        img = image.load_img(img_path, target_size=(224, 224))  # Ensure the target_size matches the input size of your model
+
+        # Preprocess the image
+        img_array = image.img_to_array(img)
+        img_array = np.expand_dims(img_array, axis=0)
+
+        # Make prediction
+        prediction = model.predict(img_array,verbose=0)
+        predicted_class = np.argmax(prediction, axis=1)[0]
+        # Assuming train_images.class_indices is a dictionary mapping class names to indices
+        class_indices = [0,1,2,3,4,5,6,7,8,9]
+        class_names =['Crane', 'Crow', 'Egret', 'Kingfisher', 'Myna', 'Peacock', 'Pitta', 'Rosefinch', 'Tailorbird', 'Wagtail']
+        predicted_class_name = class_names[predicted_class]
+
+        writer.writerow([img_file[:img_file.index('.jpg')], predicted_class_name,predicted_class])
+        qq+=1
+
+print("Predictions saved to predictions.csv")

improve.py

@@ -0,0 +1,107 @@
+import tensorflow as tf
+import os
+
+import pathlib
+import numpy as np
+data_dir = pathlib.Path("/Users/rosh/Downloads/Train_data")
+class_names = np.array(sorted([item.name for item in data_dir.glob("*")]))
+class_names = list(class_names)
+import matplotlib.pyplot as plt
+import matplotlib.image as mpimg
+import random
+# def view_random_image(target_dir, target_class):
+#
+#   target_folder = target_dir + "/" + target_class
+#
+#   random_image = random.sample(os.listdir(target_folder), 1)
+#
+#   img = mpimg.imread(target_folder + "/" + random_image[0])
+#   plt.imshow(img)
+#   plt.title(target_class)
+#   plt.axis("off")
+#
+#   print(f"Image shape: {img.shape}")
+#   plt.show()
+#   return img
+#
+#
+# #img = view_random_image(target_dir="/Users/rosh/Downloads/Train_data",target_class)
+#
+from tensorflow.keras.preprocessing.image import ImageDataGenerator
+#
+# tf.random.set_seed(42)
+#
+#
+# Define data augmentation parameters
+train_datagen = ImageDataGenerator(
+    rotation_range=20,      # Random rotation in the range [-20, 20] degrees
+    width_shift_range=0.1,  # Random horizontal shift by up to 10% of the width
+    height_shift_range=0.1, # Random vertical shift by up to 10% of the height
+    shear_range=0.2,        # Shear intensity (shear angle in radians)
+    zoom_range=0.2,         # Random zoom in the range [0.8, 1.2]
+    horizontal_flip=True,   # Random horizontal flipping
+    vertical_flip=True,     # Random vertical flipping
+    fill_mode='nearest',    # Fill mode for points outside the input boundaries
+    rescale=1./255          # Rescaling factor
+)
+
+valid_datagen = ImageDataGenerator(
+    rescale=1./255          # Rescaling factor
+ )
+#
+#
+train_dir = "/Users/rosh/Downloads/Train_data"
+#valid_dir = "/Users/rosh/Downloads/Birds"
+#
+train_data = train_datagen.flow_from_directory(directory=train_dir,
+                                               batch_size=32,
+                                               target_size=(224, 224),
+                                               class_mode="categorical",
+                                               seed=42)
+# valid_data = valid_datagen.flow_from_directory(directory=valid_dir,
+#                                                batch_size=32,
+#                                                target_size=(224, 224),
+#                                                class_mode="categorical",
+#                                                seed=42)
+
+
+# model_1 = tf.keras.Sequential([
+#     tf.keras.layers.Conv2D(filters=32, kernel_size=3, activation="relu", input_shape=(224, 224, 3)),
+#     tf.keras.layers.MaxPool2D(pool_size=2, padding="valid"),
+#     tf.keras.layers.Conv2D(64, 3, activation="relu"),
+#     tf.keras.layers.MaxPool2D(2),
+#     tf.keras.layers.Conv2D(128, 3, activation="relu"),  # Increased filters
+#     tf.keras.layers.MaxPool2D(2),
+#     tf.keras.layers.Conv2D(128, 3, activation="relu"),  # Increased filters
+#     tf.keras.layers.MaxPool2D(2),
+#     tf.keras.layers.Flatten(),
+#     tf.keras.layers.Dense(256, activation="relu"),  # Increased units
+#     tf.keras.layers.Dropout(0.5),
+#     tf.keras.layers.Dense(10, activation="softmax")
+# ])
+#
+#
+# model_1.compile(loss=tf.keras.losses.CategoricalCrossentropy(),
+#                 optimizer=tf.keras.optimizers.Adam(),
+#                 metrics=["accuracy"])
+#
+#
+# history = model_1.fit(train_data,
+#                       epochs=40,
+#                       steps_per_epoch=len(train_data),
+#                       validation_data=valid_data,
+#                       validation_steps=len(valid_data),
+#                       verbose=1)
+#
+# model_1.save("model_4.h5")
+l_model = tf.keras.models.load_model('model_4_improved_8.h5')
+
+
+l_model.compile(loss=tf.keras.losses.CategoricalCrossentropy(),
+                optimizer=tf.keras.optimizers.Adam(),
+                metrics=["accuracy"])
+l_model.fit(train_data,
+            epochs=5,
+            verbose=1)
+
+l_model.save("model_4_improved_8.h5")

test.py

@@ -0,0 +1,69 @@
+# Create a function to import an image and resize it to be able to be used with our model
+import tensorflow as tf
+import matplotlib.pyplot as plt
+import os
+import pathlib
+import numpy as np
+data_dir = pathlib.Path("/Users/rosh/Downloads/Train_data")
+class_names = np.array(sorted([item.name for item in data_dir.glob("*")]))
+class_names = list(class_names)
+class_names.pop(0)
+loaded_model = tf.keras.models.load_model('model_4_improved_8.h5')
+def load_and_prep_image(filename, img_shape=224):
+  """
+  Reads an image from filename, turns it into a tensor
+  and reshapes it to (img_shape, img_shape, colour_channel).
+  """
+  # Read in target file (an image)
+  img = tf.io.read_file(filename)
+
+  # Decode the read file into a tensor & ensure 3 colour channels
+  # (our model is trained on images with 3 colour channels and sometimes images have 4 colour channels)
+  img = tf.image.decode_image(img, channels=3)
+
+  # Resize the image (to the same size our model was trained on)
+  img = tf.image.resize(img, size = [img_shape, img_shape])
+
+  # Rescale the image (get all values between 0 and 1)
+  img = img/255.
+  return img
+
+# Adjust function to work with multi-class
+def pred_and_plot(model, filename, class_names):
+  """
+  Imports an image located at filename, makes a prediction on it with
+  a trained model and plots the image with the predicted class as the title.
+  """
+  # Import the target image and preprocess it
+  img = load_and_prep_image(filename)
+
+  # Make a prediction
+  pred = model.predict(tf.expand_dims(img, axis=0))
+
+  # Get the predicted class
+
+  pred_class = class_names[pred.argmax()] # if more than one output, take the max
+
+  # Plot the image and predicted class
+  plt.imshow(img)
+  plt.title(f"Prediction: {pred_class}")
+  plt.axis(False)
+  plt.show()
+
+pred_and_plot(loaded_model, "/Users/rosh/Downloads/egret.jpg", class_names)
+# # loaded_model.compile(loss='categorical_crossentropy',
+# #                      optimizer='adam',
+# #                      metrics=['accuracy'])
+# # Get true labels
+# valid_datagen = ImageDataGenerator(
+#     rescale=1./255          # Rescaling factor
+# )
+# valid_dir = "/Users/rosh/Downloads/Validation_data"
+# valid_data = valid_datagen.flow_from_directory(directory=valid_dir,
+#                                                batch_size=32,
+#                                                target_size=(224, 224),
+#                                                class_mode="categorical",
+#                                                seed=42)
+# pred = loaded_model.predict(valid_data)
+# preds = pred.argmax(axis=1)
+# print(preds)