Open-source ML platform for detecting deforestation, ice melt, and flooding from Sentinel-2 / Landsat imagery.

QGIS Plugin for skidtrail detection from ALS point clouds and deep learning model inference

A deep learning pipeline for semantic segmentation of cracks or defects in materials using U-Net architecture with various backbones. This project includes hyperparameter tuning, custom model architectures, and comprehensive data processing utilities.

An in-depth analysis of deep learning models U-Net and DeepLabV3+ in semantic segmentation, highlighting their applications in urban plaing, environmental monitoring, and geographic information systems.

This is a joint project of Vinit Firke (myself) and Narunat Pantapalin (https://github.com/narupanta) for the Fortgeschrittene Praktikum Data Science course happened in Winter Semester 2024-25 at TU Braunschweig.

Plataforma inteligente

End-to-end deep learning pipeline for melanoma detection using U-Net segmentation and EfficientNet classification.

This project focuses on the detection and analysis of UHIs in Hamburg.The core of this project is to leverage multi-source data, including satellite imagery and vector data, with deep learning techniques (specifically U-Net architectures) for high-resolution UHI mapping and analysis.

Deep learning model for automatic lung segmentation from chest X-ray images using U-Net architecture with ResNet34 encoder.

Binary building segmentation on aerial images using the Massachusetts Buildings dataset

U-Net for MRI segmentation.

Satellite imagery analysis using deep learning

A project for segmenting buildings in satellite images using the U-Net architecture. Includes data preprocessing, model training, and evaluation scripts, along with preprocessed datasets and trained models.

A fully automated, high-fidelity, and multi-input GAN-based framework for precise cloud removal from satellite images with enhanced interpretability and performance across varying cloud types and terrains.

A comprehensive deep learning project for detecting and segmenting brain diseases, particularly tumors, in MRI scans using multiple state-of-the-art architectures including U-Net and Meta's Segment Anything Model (SAM).