├── File 3
├── README.md
└── file1


/File 3:
--------------------------------------------------------------------------------
 1 | import os
 2 | import pandas as pd
 3 | 
 4 | def extract_image_metadata_to_excel(data_dir='data', output_file='plastic_waste_dataset.xlsx'):
 5 |     records = []
 6 | 
 7 |     for label in ['plastic', 'non_plastic']:
 8 |         class_dir = os.path.join(data_dir, label)
 9 |         for filename in os.listdir(class_dir):
10 |             if filename.endswith(".png"):
11 |                 records.append({
12 |                     'file_name': filename,
13 |                     'label': label,
14 |                     'file_path': os.path.join(class_dir, filename)
15 |                 })
16 | 
17 |     df = pd.DataFrame(records)
18 |     df.to_excel(output_file, index=False)
19 |     print(f"Dataset metadata saved to {output_file}")
20 | 
21 | if __name__ == "__main__":
22 |     extract_image_metadata_to_excel()
23 | 


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
 1 | # AI for Automated Classification of Plastic Waste in Ocean Images
 2 | 
 3 | This project uses synthetic data to simulate ocean images containing plastic waste and trains a Convolutional Neural Network (CNN) to classify them.
 4 | 
 5 | ## Features
 6 | 
 7 | - Synthetic data generator
 8 | - Binary image classification (plastic vs. non-plastic)
 9 | - CNN built using TensorFlow/Keras
10 | - Training, evaluation, and prediction
11 | 
12 | ## Setup
13 | 
14 | ```bash
15 | pip install tensorflow matplotlib numpy
16 | python plastic_classifier.py
17 | 
18 | Predicting a New Image
19 | 
20 | Uncomment and run:
21 | 
22 | predict_image('data/plastic/plastic_1.png')
23 | 
24 | Directory Structure
25 | 
26 |     data/: Auto-generated synthetic images
27 | 
28 |     model/: Saved model weights
29 | 
30 | 
31 | ---
32 | 
33 | Would you like to upgrade the project with:
34 | - real image augmentation?
35 | - Streamlit dashboard for visual interface?
36 | - Deployment tips (e.g., Flask, FastAPI, or Hugging Face)?
37 | 
38 | Let me know how far you want to go with this!
39 | 


--------------------------------------------------------------------------------
/file1:
--------------------------------------------------------------------------------
 1 | import os
 2 | import numpy as np
 3 | import matplotlib.pyplot as plt
 4 | import tensorflow as tf
 5 | from tensorflow.keras.models import Sequential
 6 | from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout
 7 | from tensorflow.keras.preprocessing.image import ImageDataGenerator, array_to_img
 8 | 
 9 | # STEP 1: Generate synthetic data (if data doesn't exist)
10 | def generate_synthetic_data(base_dir='data', num_samples=500, img_size=(128, 128)):
11 |     os.makedirs(base_dir, exist_ok=True)
12 |     for category in ['plastic', 'non_plastic']:
13 |         category_path = os.path.join(base_dir, category)
14 |         os.makedirs(category_path, exist_ok=True)
15 |         for i in range(num_samples):
16 |             img = np.random.rand(*img_size, 3)  # Random noise image
17 |             if category == 'plastic':
18 |                 img[30:60, 30:60] = [1.0, 1.0, 0.0]  # Simulate yellowish floating object
19 |             else:
20 |                 img *= 0.5  # Darker background
21 | 
22 |             img = (img * 255).astype(np.uint8)
23 |             plt.imsave(f'{category_path}/{category}_{i}.png', img)
24 | 
25 | # STEP 2: Data preparation
26 | def prepare_data(data_dir='data', img_size=(128, 128), batch_size=32):
27 |     datagen = ImageDataGenerator(rescale=1./255, validation_split=0.2)
28 | 
29 |     train_gen = datagen.flow_from_directory(
30 |         data_dir,
31 |         target_size=img_size,
32 |         batch_size=batch_size,
33 |         class_mode='binary',
34 |         subset='training'
35 |     )
36 | 
37 |     val_gen = datagen.flow_from_directory(
38 |         data_dir,
39 |         target_size=img_size,
40 |         batch_size=batch_size,
41 |         class_mode='binary',
42 |         subset='validation'
43 |     )
44 | 
45 |     return train_gen, val_gen
46 | 
47 | # STEP 3: CNN model
48 | def build_model(input_shape=(128, 128, 3)):
49 |     model = Sequential([
50 |         Conv2D(32, (3, 3), activation='relu', input_shape=input_shape),
51 |         MaxPooling2D(2, 2),
52 |         Conv2D(64, (3, 3), activation='relu'),
53 |         MaxPooling2D(2, 2),
54 |         Flatten(),
55 |         Dense(128, activation='relu'),
56 |         Dropout(0.3),
57 |         Dense(1, activation='sigmoid')
58 |     ])
59 |     model.compile(optimizer='adam',
60 |                   loss='binary_crossentropy',
61 |                   metrics=['accuracy'])
62 |     return model
63 | 
64 | # STEP 4: Train and Save Model
65 | def train_model():
66 |     generate_synthetic_data()
67 |     train_gen, val_gen = prepare_data()
68 |     model = build_model()
69 |     model.summary()
70 |     model.fit(train_gen, epochs=5, validation_data=val_gen)
71 |     os.makedirs('model', exist_ok=True)
72 |     model.save('model/plastic_classifier.h5')
73 | 
74 | # STEP 5: Prediction on new image
75 | def predict_image(img_path, model_path='model/plastic_classifier.h5', img_size=(128, 128)):
76 |     model = tf.keras.models.load_model(model_path)
77 |     img = tf.keras.preprocessing.image.load_img(img_path, target_size=img_size)
78 |     img_array = tf.keras.preprocessing.image.img_to_array(img) / 255.0
79 |     img_array = np.expand_dims(img_array, axis=0)
80 |     prediction = model.predict(img_array)[0][0]
81 |     label = "Plastic" if prediction > 0.5 else "Non-Plastic"
82 |     print(f"Prediction: {label} ({prediction:.2f})")
83 | 
84 | if __name__ == "__main__":
85 |     train_model()
86 |     # Example prediction
87 |     # predict_image('data/plastic/plastic_1.png')
88 | 


--------------------------------------------------------------------------------