360 Field Campaign
Collecting panoramic video across the Midwest for façade understanding and civic infrastructure monitoring.
Highlighted areas
TRACE
TRACE (Transformer with Reinforcement and Actor-Critic for Edge-vectorization) fuses DeepLab-UNet segmentation with an actor–critic agent that incrementally grows vector networks, producing complete, routable road graphs from aerial and satellite imagery.
Repository access is restricted ( TRACE GitLab ). Contact the lab for credentials.
LAMDA
LAMDA (Low-shot Algorithm for Mine Data Augmentation) develops low-shot learning and generative augmentation strategies for automatic target detection when data are scarce or classified, improving remote-sensing models for rare or hazardous objects.
Repository access is restricted ( LAMDA GitLab ). Contact the lab for credentials.
Architectural Facade Mining
Multi-source imagery (Flickr, panoramic extracts) for window detection, stylistic classification, and metadata fusion.
LiteLLM GPU Stack
1080 Ti cluster that mirrors the OpenAI API so students can prototype without commercial credits.