This project focuses on Automated Land Cover Classification (LCC) using Deep Learning. By implementing a ResNet-50 Convolutional Neural Network, the model successfully classifies high-resolution satellite imagery into 10 distinct environmental categories. This mirrors industry-standard workflows used by organizations such as NRSC/ISRO for geospatial analysis.
- Backbone: ResNet-50 (Transfer Learning from ImageNet)
- Data Pipeline: Customized
torchvisiontransforms including normalization and random flips. - Classification Head: Fully connected layers with Dropout (0.3) to prevent overfitting.
- Optimization: Adam Optimizer with Negative Log-Likelihood Loss (NLLLoss).
The model was trained on the EuroSAT (RGB) dataset, derived from Sentinel-2 satellite imagery. Classes include: Annual Crop, Forest, Herbaceous Vegetation, Highway, Industrial, Pasture, Permanent Crop, Residential, River, Sea Lake.
- Accuracy: ~96% on Validation Set.
- Efficiency: Leveraged GPU acceleration (T4) in Colab for fast convergence.
- Evaluation: Performance verified via Confusion Matrix and F1-score analysis.
- Install the required libraries:
pip install torch torchvision matplotlib seaborn scikit-learn tqdm