PRESS PLAY
The official podcast of IQT CosmiQ Works, an applied research lab dedicated to emerging geospatial analytics
At SpaceNet, a road-detection score gets interesting when you ask whether its roads actually support a route.
Press play if you want
Short reviews from the PlayNext desk, based on the episodes we processed.
A farewell reunion on geospatial AI, research culture, and choosing metrics around the failures that matter.
Training_Data's #29 treats Cosmic Works' anticipated final podcast as a reunion for the people shaped by it, while Solaris and SpaceNet give the episode its technical backbone.
A grounded review of how change detection can support disaster planning, response, and mapping.
SpaceNet 7 treats change detection over time as a practical tool for disaster planning, response, and mapping. Grace Kitzmiller, Principal Product Manager at AWS Disaster Response, ties the capability to flooding impacts, resource planning, faster response, and signals that…
A measured look at how SpaceNet 6 frames SAR, optical imagery, and open geospatial research.
SpaceNet 6 works best here as a focused case study in geospatial AI, especially where SAR and optical imagery meet. With Jake Schirmeier filling in for Ryan Lewis, Daniel Hogan and Ronnie Hinch walk through the challenge baseline, the appeal of SAR data fusion, and the practical…
SpaceNet 6 gives a technical mapping challenge shape through concrete scores, margins, and competitors.
SpaceNet 6 turns a specialized mapping problem into a focused look at multi-sensor computer vision, using co-registered electro-optical and synthetic aperture radar data as the center of the challenge.
A clear review of how SpaceNet 5 judged road-network models once speed became part of the score.
SpaceNet 5 receives a practical, results-minded review built around road-network detection and optimized routing. Adam Van Etten walks through the released dataset, the baseline approach, and the research that followed, then broadens the lens to show how challenge results can be…
AI research, explained by the people doing it.
Talking shop with the people behind data products
Specialists walking through software's tradeoffs, one system at a time.
Founders and engineers on AI coding agents like Claude Code, and supply chain security.
Anna Van Etten hosts a final Training Data reunion with Ryan Lewis, Lisa Porter, Daniel Hogan, Jake Schirmeier, Nick Weir, and Dave Lindenbaum's work repeatedly referenced.
Training_Data episode #28 centers on the SpaceNet 7 Challenge, which moves the series from static building footprints to multi-temporal urban development.
["Training Data's SpaceNet 6 episode brings Jake Schirmeier, Daniel Hogan, Ronnie Hinch, and Jason Brown together around Capella SAR and Maxar WorldView-2 imagery over Rotterdam, Netherlands.", "Daniel says the baseline used transfer learning from optical data into SAR training, and he reports that the optical-pretrained model beat a SAR-only model.", "Jason and Ronnie walk through HSV pan-sharpening, SAR-to-optical GAN translation, the limits of regression, and why SAR features such as layover, speckle, and backscattering are hard to read directly.", "The later sections cover 15 average revisits, a 20% performance gain at four looks, Solaris' Pipe Segment API, and the planned release of all 204 original complex Rotterdam strips."]
["Training Data's SpaceNet 6 results episode centers on Rotterdam, the Netherlands, and a co-registered SAR-plus-optical challenge for building-footprint extraction.", "Ryan talks with Jake Schirmeier and Daniel Hogan about why SAR matters when optical imagery is unavailable and why the test phase used SAR-only inputs.", "Daniel's UNET-with-VGG11 baseline starts around 0.21 F1 under the SpaceNet metric, while the winning model reaches 42.4.", "Jake says the challenge drew more than 1,600 submissions, the top models leaned on EfficientNet and fast training, and future work includes colorization, city-scale evaluation, and complex SAR releases."]
Training_Data episode #25 covers SpaceNet 5, the road network detection and optimized routing challenge.
An interview with Carto's founder and director of spatial data science about making geospatial analytics more accessible to various users and business verticals.
A discussion on structuring businesses around open-source software in the geospatial analytics market.
This episode discusses the SpaceNet collaboration and upcoming competition focused on road networks and routing using satellite imagery.
An interview with Peter Wang, CTO of Anaconda, discussing data science and geospatial analytics.