Terrapattern – Neural network visual search tool for satellite imagery

Category Javascript Members Processing People , , Posted on:

Created by Golan Levin, David Newbury, and Kyle McDonald, with the assistance of Golan's students at CMU, Irene Alvarado, Aman Tiwari, and Manzil Zaheer, Terrapattern is a visual search tool for satellite imagery. The project provides journalists, citizen scientists, and other researchers with the ability to quickly scan large geographical regions for specific visual features.

Terrapattern uses a deep convolutional neural network (DCNN), based on the ResNet ("Residual Network") architecture developed by Kaiming He et al. The team trained a 34-layer DCNN using hundreds of thousands of satellite images labeled in OpenStreetMap, teaching the neural network to predict the category of a place from a satellite photo. In the process, their network learned which high-level visual features (and combinations of those features) are important for the classification of satellite imagery. They used 466 of the Nominatim categories (such as "airport", "marsh", "gas station", "prison", "monument", "church", etc.), with approximately 1000 satellite images per category. Their resulting model, which took 5 days to compute on an nVidia 980 GPU, has a top-5 error rate of 25.4%.

Sorry, this is Members Only content. Please Log-in.

Join us today by becoming a Member.

  • Archive: More than 3,500 project profiles, scores of essays, interviews and reviews.
  • Publish: Post your projects, events, announcements.
  • No Ads: No advertisements, miners, banners.
  • Education: Tutorials (beginners and advanced) with code examples, downloads.
  • Jobs Archive: Find employers who have recruited here in the past (over 800 jobs).
  • Discounts: Special offers and giveaways (events, books and media).

Become a Member

/++

Designing the Computational Image, Imagining Computational Design – An Excavation of Digital Form

“Designing the Computational Image, Imagining Computational Design” is an exhibition that excavates the foundation of computer-aided design and manufacturing and weaves together several ‘origin stories’ for contemporary consideration. The show recently closed after a seven-week run at the Miller Gallery at Carnegie Mellon University in Pittsburgh, and CAN was fortunate enough to get a guided tour with curator Daniel Cardoso Llach as it was winding down.

Blackberry Winter – Artificial human motion in asymmetry

Created by Christian Mio Loclair (Waltz Binaire), ‘Blackberry Winter’ is an investigation into the possibilities of identify motion as a continuous walk in a latent space of situations.

Constructive Interference – Hypersonic & Plebian Design

Created as a collaboration between Hypersonic and Plebian Design, Constructive Interference is a sculpture designed to engage passer-bys using the wonder of moire patterns. Installation is composed of two large patterned sheets of steel, designed to create a rapidly changing visual interference effect as viewers pass by.

Him & Her (Interconnection)

A series of three pictures exploring tension and harmony. Each unique edition was created using algorithms, calculating each pixel in relation to the next one. An interplay between tension and harmony, opposing each other in the first two editions and finally merging into a beautiful complexity in the third artwork “Interconnection”. Available as exclusive NFTs…

/+

/++

ScreenLab 0x02 – Exploring new modes of perception

Sorry, this is Members Only content. Please Log-in. Join us today by becoming a Member. Archive: More than 3,500 project profiles, scores of essays, interviews and reviews.Publish: Post your projects, events, announcements.No Ads: No advertisements, miners, banners.Education: Tutorials (beginners and advanced) with code examples, downloads.Jobs Archive: Find employers who have recruited here in the past…