Over the past 6 years we've helped the intelligence & academic communities better understand how #geospatial machine learning can address hard problems. As CosmiQ Works is folded into #IQTLabs, our final blog highlights accomplishments from over the years! ow.ly/EUvg50E3rfC
Over the past 6 years we've helped the intelligence and academic communities better understand how #geospatial machine learning can help tackle hard problems. Today's podcast brings together current colleagues and alum to celebrate key accomplishments! ow.ly/iRCO50DUgbU
Building footprints provide a useful proxy for many humanitarian applications. Our new paper discusses efforts to develop techniques for precise building footprint localization, tracking, & change detection via the #SpaceNet7 challenge. ow.ly/lkjm50DL8lI#geospatial
Building tracking and change detection from moderate satellite imagery is surprisingly tractable. Our final #SpaceNet7 analysis blog explores winning model performance with an interactive map and provides details on how to access the #opensource models. ow.ly/cAiq50DBTpX
The new @Spacenet_AI blog analyzes results of #SpaceNet7 to explore how the SpaceNet Change and Object Tracking (SCOT) metric acts with real-world data. Do different parts of the metric really measure what they claim to measure? Spoiler alert - they do! ow.ly/CF9a50DuBDk
What was a main theme in the #SpaceNet7 challenge results? Overachieving pixels. Our new blog dives into this, comparing results to past SpaceNet challenges, and explores predictions with an interactive map. ow.ly/Zb2F50Dm0VY#geospatial#opensource
Part II of our #SpaceNet7 results deep dive is here! Learn more about the winning algorithm, the innovations that vastly improved upon the baseline model, and an unexpected performance curve. ow.ly/NLFZ50COxq4 Stay tuned next month as we continue to explore results.
Results are in from the #SpaceNet7 challenge! Our first challenge evaluation blog details the winners and explores insights gleaned from the challenge. Results will also be featured at #NeurIPS2020! ow.ly/OLWV50CAO1T#geospatial#opendata
Interested in learning how to build an open source analytics ecosystem for #geospatial applications? Tune into our session from #IGARSS2020 last month where we detail how we did this through @Spacenet_AI data ow.ly/IAF350CtP9W
Interested in applied machine learning for geospatial applications? Join CosmiQ's Adam Van Etten (ow.ly/eEM450CbopO) & Jake Shermeyer (ow.ly/E6yn50CbopM) as they present at MLconf 2020 this Friday. Check out the agenda and register here: mlconf.com
@AdamVanEtten's talk focuses on the City-Scale Road Extraction from Satellite Imagery (CRESI) algorithm (ow.ly/zcG050CboDk) that rapidly extracts road networks and identifies speed limits and route travel times for each roadway, using only satellite imagery as input
@JakeShermeyer 's talk is on the openly licensed RarePlanes (ow.ly/vS7l50CboMC) dataset & research study that examines the value of @aireverie synthetic data to aid CV algorithms in their ability to automatically detect aircraft and their attributes in satellite imagery.
Interested in attending another great talk at #IGARSS2020? Check out our work on road network extraction from multiple satellite look angles with @Spacenet_AI data. The talk will be live this Friday at 9:25AMET! #geospatial#opendataow.ly/tcDx50BFn5r
We've released the expanded #SpaceNet6 dataset containing 202 untiled @capellaspace SAR image strips & a @Maxar WorldView 2 scene! The SAR data is the 1st openly licensed quad-polarized imagery at 0.25m spatial resolution that spans such an extensive area. ow.ly/Ts4C50Bz5Ah
We wish the best of luck to our colleague @RyanSLewis07 on his last day with us as he pursues his next career endeavor! CosmiQ never would have been born without him, and his leadership and creativity has significantly contributed to how far we've come today.
Our Cloud Native Geospatial Outreach Day & Sprint is one week away! We're working to welcome to anyone who is interested - no geospatial or software expertise required. See medium.com/@cholmes/welcomin… for more info. Please share widely so we can bring in new users & collaborators!