CRATERS

import os #read/write in the local file system

import cv2 #image processing

import matplotlib.pyplot as plt #x,y-axis plotting

import pandas as pd #data manupulation and analysis library (handles data - update,read)

import numpy as np #mathematical operations

import plotly.express as px #mathematical plot - using images

import plotly.graph_objects as go # ''

import seaborn as sns #data visualization

from sklearn.model_selection import train_test_split #train and test data splitups

train_img_path = "/content/drive/MyDrive/Craters project/train/images"

train_lbl_path = "/content/drive/MyDrive/Craters project/train/labels"

valid_img_path = "/content/drive/MyDrive/Craters project/valid/images"

valid_lbl_path = "/content/drive/MyDrive/Craters project/valid/labels"

test_img_path = "/content/drive/MyDrive/Craters project/test/images"

test_lbl_path = "/content/drive/MyDrive/Craters project/test/labels"

model_path = "/content/drive/MyDrive/Craters project/best.pt"

data_yaml_path = "/content/drive/MyDrive/Craters project/data.yaml"

# image definition and identification (like: identifying as male or female)

def load_labels(label_path):

label_files = os.listdir(label_path)

data = []

classes = set()

for file in label_files:

with open(os.path.join(label_path, file), 'r') as f:

lines = f.readlines()

for line in lines:

parts = list(map(float, line.strip().split()))

data.append([file, *parts])

classes.add(int(parts[0]))
 df = pd.DataFrame(data, columns=['file', 'class', 'x_center', 'y_center', 'width', 'height'])

return df, sorted(classes)

train_labels, train_classes = load_labels(train_lbl_path)

valid_labels, valid_classes = load_labels(valid_lbl_path)

test_labels, test_classes = load_labels(test_lbl_path)

# image seperation for training, testing and valid class (11,000 - total images added)

# class[0] - has image

# class[1,2,3,4,5....] - do not have image

all_classes = sorted(set(train_classes + valid_classes + test_classes))

class_names = [f'class_{i}' for i in all_classes]

# Output for the above cell

print("Train Labels")

print(train_labels.head())

print("\nValidation Labels")

print(valid_labels.head())

print("\nTest Labels")

print(test_labels.head())

# storing in a file extension .yaml

# Why? becz of Yolo model(deep learning)

data_yaml_content = f"""

train: {train_img_path}

val: {valid_img_path}

test: {test_img_path}

nc: {len(all_classes)} # number of classes

names: {class_names} # class names

"""

with open(data_yaml_path, 'w') as f:

f.write(data_yaml_content)
# Plotting distribution of bounding box sizes

# making a box like identification in craters

# classified based on area

def plot_bounding_box_distribution(labels, title):

labels['area'] = labels['width'] * labels['height']

fig = px.histogram(labels, x='area', nbins=50, title=title)

fig.show()

plot_bounding_box_distribution(train_labels, 'Train Bounding Box Area Distribution')

plot_bounding_box_distribution(valid_labels, 'Validation Bounding Box Area Distribution')

plot_bounding_box_distribution(test_labels, 'Test Bounding Box Area Distribution')

# Image Preprocessing and Visualization

# preprocessing: removing the unwanted data

# bgr to rgb: to process the data and vizualization purpose

def visualize_sample_images(image_path, label_df, n_samples=5):

image_files = os.listdir(image_path)[:n_samples]

for img_file in image_files:

img_path = os.path.join(image_path, img_file)

img = cv2.imread(img_path)

img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

fig, ax = plt.subplots(1, 1, figsize=(10, 10)) # standard size to display

ax.imshow(img)

labels = label_df[label_df['file'] == img_file]

for _, label in labels.iterrows():

x_center = int(label['x_center'] * img.shape[1])

y_center = int(label['y_center'] * img.shape[0])

width = int(label['width'] * img.shape[1])

height = int(label['height'] * img.shape[0])

x_min = x_center - width // 2

rect = plt.Rectangle((x_min, y_min), width, height, edgecolor='red', facecolor='none',

ax.add_patch(rect)

plt.title(f'Sample Image: {img_file}')

plt.axis('off')

plt.show()

visualize_sample_images(train_img_path, train_labels)

visualize_sample_images(valid_img_path, valid_labels)

visualize_sample_images(test_img_path, test_labels)

# You Only Look Once

# Detecting an specific object from the image

!pip install ultralytics

from ultralytics import YOLO

model = YOLO('yolov8n.pt')

model.train(data=data_yaml_path, epochs=10)

# Visualize sample detections

# Created without label

def visualize_detections(model, image_path, n_samples=10):

image_files = os.listdir(image_path)[:n_samples]

for img_file in image_files:

img_path = os.path.join(image_path, img_file)

img = cv2.imread(img_path)

img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

results = model(img_path)

fig, ax = plt.subplots(1, 1, figsize=(10, 10))

ax.imshow(img)
 for result in results[0].boxes:

x_min, y_min, x_max, y_max = result.xyxy[0].tolist()

conf = result.conf[0].item()

rect = plt.Rectangle((x_min, y_min), x_max - x_min, y_max - y_min, edgecolor='red',

ax.add_patch(rect)

ax.text(x_min, y_min, f'{conf:.2f}', bbox=dict(facecolor='yellow', alpha=0.5))

plt.title(f'Detection in: {img_file}')

plt.axis('off')

plt.show()

visualize_detections(model, test_img_path)

print("Model training, evaluation, and sample visualization completed. The trained model is saved at
'/kaggle/working/best_model.pt'.")