cuvis.sdk
- 24
- Python
- Apache-2.0
- 5 d ago
cubert-hyperspectral
Developer of Hyperspectral Cameras and Software Solutions
Built from cubert-hyperspectral's public GitHub profile, Ulm, Germany. Not on Take My Repo yet, so there are no posts or updates here.
Cuvis.AI is an open-source, extensible framework for building modular AI pipelines that process, analyze, and deploy hyperspectral video data in real-time applications.
Core framework for cuvis-ai: Node base class, pipeline graph, typed ports, gRPC API, training orchestration, and plugin system
CentroidNetV2 nodes for cuvis-ai: counting and instance segmentation of small, touching objects by centroid and border voting
SteerViT for cuvis-ai: prompt-steered DINOv2 patch features and a zero-shot prompted anomaly map
HSI-PatchCore for cuvis-ai: memory-bank anomaly detection on spectra or dense feature grids, score calibration and fusion
Configurable U-Net (DynUNet) nodes for cuvis-ai: 2D, factorised 2.5D, and 3D hyperspectral segmentation, plus Dice and cross-entropy losses
RF-DETR (Roboflow, Apache-2.0 tier) object detection as a cuvis.ai plugin node
Pluggable hyperspectral DataModules (cu3s + COCO, TIFF + paired-PNG) for the Cuvis.AI ecosystem
Cuvis.AI reproduction of the InSpecScrap SWIR hyperspectral steel-scrap classifier (Gursch et al. 2026)
Data-augmentation plugin for cuvis.ai hyperspectral training pipelines