Synthetic DSM Generation with Stable Diffusion
Master thesis with DLR — generative CV models that synthesize Digital Surface Models from DTM and building-mask inputs.
Overview
Master thesis in collaboration with DLR (German Aerospace Center). The model takes a Digital Terrain Model and a building mask as conditioning inputs and generates a Digital Surface Model (DSM). Built on top of Stable Diffusion, trained and benchmarked end-to-end with custom evaluation harnesses for geospatial fidelity.
Key Highlights
- Conditioned Stable Diffusion on DTM + building mask inputs
- Custom evaluation pipeline for geospatial output fidelity
- Collaboration with DLR Earth Observation Center
Tech Stack
Other Projects
All ProjectsProjection Mapping in Unity3D
Spatial calibration + blending for multi-projector image/video content, in Unity3D.
View ProjectArc-Mode Classification + PyQt GUI
Deep-learning pipeline that classifies arc appearance modes from high-speed video, with a PyQt analysis GUI.
View ProjectMultitask NLP Classifier with BERT
Multitask BERT for paraphrase detection, semantic textual similarity and sentiment analysis.
View ProjectStay Updated
Subscribe to get the latest blog posts, project updates, and data science insights straight to your inbox.
No spam. Unsubscribe anytime.