Scribd
Upload
30 views

LULC Classification Presentation

The document outlines a deep learning approach using Convolutional Neural Networks (CNN) for classifying satellite images into land use/land cover categories using the EuroSAT dataset. It details the data processing, augmentation techniques, CNN model architecture, training process, and evaluation methods, highlighting the model's high accuracy and potential future improvements. Suggestions for future work include experimenting with deeper architectures and integrating multi-temporal data for enhanced classification.

Uploaded by

EC-223 Shivakumar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
30 views

LULC Classification Presentation

The document outlines a deep learning approach using Convolutional Neural Networks (CNN) for classifying satellite images into land use/land cover categories using the EuroSAT dataset. It details the data processing, augmentation techniques, CNN model architecture, training process, and evaluation methods, highlighting the model's high accuracy and potential future improvements. Suggestions for future work include experimenting with deeper architectures and integrating multi-temporal data for enhanced classification.

Uploaded by

EC-223 Shivakumar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 8

Land Use Land Cover (LULC)

Classification Using CNN


A Deep Learning Approach for
Satellite Image Classification
Introduction
• - Objective: Classify satellite images into land
use/land cover categories.
• - Dataset: EuroSAT (Images of different land
types such as forests, urban areas, etc.).
• - Approach: Use Convolutional Neural
Networks (CNN) for classification.
Data Processing
• - Loaded images from dataset and resized to
64x64 pixels.
• - Normalized pixel values (0-1) for better
training stability.
• - One-hot encoded labels for multi-class
classification.
• - Split dataset into 80% training and 20%
testing.
Data Augmentation
• - Used ImageDataGenerator to improve model
generalization.
• - Applied transformations:
• - Rotation (20°)
• - Width & height shift (20%)
• - Horizontal flip
• - Helps reduce overfitting and enhances
recognition.
CNN Model Architecture
• - Convolutional Layers (32, 64, 128 filters) with
ReLU activation.
• - Batch Normalization for stability.
• - MaxPooling for feature extraction.
• - Fully connected layer (256 neurons) with
Dropout (50%).
• - Softmax activation for multi-class
classification.
Model Training
• - Optimizer: Adam (learning rate = 0.001).
• - Loss function: Categorical Crossentropy.
• - Training for 10 epochs using augmented
data.
• - Callbacks: ReduceLROnPlateau &
EarlyStopping to prevent overfitting.
Model Evaluation
• - Accuracy: Evaluated on test data.
• - Confusion Matrix: Visualizing correct &
incorrect predictions.
• - Precision-Recall Curve: Assessing per-class
performance.
• - Training history graphs: Accuracy & loss
trends over epochs.
Results & Future Work
• - The model achieves high accuracy in
classifying LULC images.
• - Possible Improvements:
• - Experiment with deeper architectures
(ResNet, EfficientNet, etc.).
• - Use larger image resolutions for better
feature extraction.
• - Integrate multi-temporal data for change
detection.

You might also like