# Project Title: CNN for Detecting Deforestation and Ice Melting

## Project Description

This project investigates the use of Convolutional Neural Networks (CNNs) to detect deforestation and ice melting from satellite imagery. Utilizing state-of-the-art CNN architectures, students will develop models capable of identifying and monitoring environmental changes over time. These models can be essential tools for environmental monitoring agencies and researchers in understanding and mitigating the impacts of climate change. For reference, you can explore similar studies here:
https://www.mdpi.com/2072-4292/12/6/901
https://ieeexplore.ieee.org/abstract/document/8900323/

## Project Steps

### Part 1: Data Collection and Preliminary Analysis (10 points)

#### Data Collection
- Obtain datasets of satellite images depicting forested and ice-covered areas from sources such as NASA Earth Observing System Data and Information System (EOSDIS) or Sentinel Hub (https://eos.com/).

#### Data Cleaning
- Clean the datasets by removing corrupted images, ensuring consistent resolution and format, and annotating images for deforestation and ice melting events.

#### Preliminary Data Analysis and Visualization
- Perform preliminary analysis to identify patterns and anomalies in the satellite imagery.
- Visualize the data using geographical maps for spatial distribution.

### Part 2: Deep Learning Model (10 points)

#### Build CNN Model
- Develop a CNN model using PyTorch. Incorporate layers suitable for image classification and segmentation tasks.
- Provide a detailed walkthrough of the code, which should be well-commented and available in a repository linked in the report.

#### Model Training and Validation
- Train the model on the prepared dataset, using a split of training and validation data to monitor for overfitting.
- Validate the model's performance using appropriate metrics such as accuracy, precision, recall, and F1-score. Adjust hyperparameters as needed to improve outcomes.

#### Deployment and Real-Time Testing
- Deploy the model to perform real-time detection of deforestation and ice melting from incoming satellite images.
- Visualize the results and assess the model's effectiveness in real-world scenarios.

### Part 3: Environmental Impact Analysis (10 points)

#### Analysis Using the Predictive Model
- Analyze the results obtained from the CNN model to understand trends in deforestation and ice melting.
- Evaluate the implications of these trends on global climate patterns and local ecosystems.

#### Presentation and Report
- Summarize the findings and methodologies in a comprehensive report. Reflect on the model's efficacy and areas for improvement.
- Discuss potential societal impacts, including the benefits and challenges of using AI for environmental monitoring.

## Grading Criteria

### Presentation of Data and Preliminary Data Analysis (10 points)
- Proper handling and presentation of the dataset.
- Clarity and relevance of the visualizations.
- Thoroughness of the initial analytical findings.

### Deep Learning Model (10 points)
- Accuracy and robustness of the CNN model.
- Model architecture design and implementation.
- Detailed documentation and reproducibility of the model training process.

### Environmental Impact Analysis (10 points)
- Depth and rigor of the analysis regarding the model's real-world applicability.
- Discussion of ethical considerations and potential biases.