SBIR-STTR Award

Kraenion is an AI company developing Machine Learning and Active Learning technology to an-alyze large 2D and 3D multi-spectral datasets of importance to public safety and national secu-rity. Our deep learning models and statistical vision algorithms process planet-scale satellite image datasets and security critical CT and X-ray imagery. Kraenion’s Vision Engine platform includes active learning based neural network training technology where the training software is aware of the cost of labeling data. Unlike traditional neural network training that assumes a large labeled dataset, our system carefully picks samples that maximize the learning opportunity and presents it for labeling to a human annotator. This provides much higher return on dollars invested for data annotation in areas like satellite imagery where unlabeled data is abundant, but there is a scarcity of labeled data. In this SBIR, we propose to extend and adapt our innovations in deep learning and active learning to a) Risk assessment based on a combination of satellite/airborne imagery and ancillary GIS data such as maps of the electric grid and municipal building permits. b) NOAA applications including coastal change analysis, land cover mapping and the annotation of weather data for Earth System Models.

Kraenion is an AI company developing Machine Learning and Active Learning technology to analyze large 2D and 3D multi-spectral datasets of importance to public safety and national security. Our deep learning models and statistical vision algorithms process planet-scale satellite image datasets and security critical CT and X-ray imagery. Kraenion’s Vision Engine platform includes active learning based neural network training technology where the training software is aware of the cost of labeling data. Unlike traditional neural network training that assumes a large labeled dataset, our system carefully picks samples that maximize the learning opportunity and presents it for labeling to a human annotator. This provides much higher return on dollars invested for data annotation in areas like satellite imagery where unlabeled data is abundant, but labeled data is scarce. We propose to extend our innovations in deep learning and active learning to a) Risk assessment based on a combination of satellite/airborne imagery and ancillary GIS data such as maps of the electric grid and municipal building permits. b) Active learning based secure image annotation and ML technologies for federal agencies involved in satellite image analysis. Potential federal customers include NASA, NOAA, NGA, USGS, USDA and DHS.