CO2 Emissions Segmentation from Satellite Images

•Developed a deep learning-based system for segmenting CO2 emission regions from satellite

images with 13-channel spectral data (.tif format).

•Utilized and fine-tuned DeepLab models to achieve pixel-level segmentation with high accuracy,

leveraging their capability for capturing multi-scale context.

•Preprocessed large-scale satellite data, handling multi-channel images efficiently and normalizing

spectral bands to enhance feature extraction.

•Designed an end-to-end pipeline for data ingestion, segmentation, and visualization of results,

enabling actionable insights into emission hotspots.