Aquanty Hosts Machine Learning Challenge: Enhancing Satellite Image Mosaics
Aquanty is excited to sponsor a machine learning coding / product development challenge, hosted by our friends at Aggregate Intellect. If you are a data scientist, software engineer or a professional machine learning engineer then read on – your team can win a $5,000 prize!
What is the challenge?
Aquanty operates a web platform called AgSat, which primarily serves satellite imagery and geospatial data to end-users in the agricultural sector. We are looking to assemble a mosaic of frequently updated, high-resolution/sharp-looking images that can be used as a basemap within the AgSat platform.
Currently, only archived high-resolution images for some areas are available (at ~1 m resolution); they are broadly representative, but not up-to-date. High-frequency images with good coverage are available, but at much lower resolution than desirable for our application. In addition, high-frequency data is often further contaminated by the presence of clouds in the image.
We would like to have a system that can generate images with high apparent resolution from the high-frequency/low-resolution data, which can be used to assemble a frequently updated, cloud-free, and continuous mosaic.
Figure 1: The Goal – Enhanced Satellite Image Mosaics, image source: https://remicres.github.io/super-resolution (Interactive Urban Demo)
Data Sources
We anticipate that the Sentinel-2 satellites will be the main source of data for this machine learning exercise. Sentinel-2 satellites have a resolution of 10 m, an update cycle of 5 days and multiple spectral bands. The goal would be a module that can ingest low-resolution satellite imagery, ideally from the Sentinel-2 satellites, and generate realistic-looking images that are free of clouds, largely free of artifacts or errors and have an apparent resolution on the meter-scale.
Aquanty is sponsoring this challenge through the ‘Machine Learning in Climate and Sustainability’ cohort on the Aggregate Intellect platform. To learn more about the challenge visit the Aggregate Intellect website and review Cohort 9 – Deep Learning Products challenge.
What is AgSat, and how does this competition fit in?
AgSat (www.AgSat.ca) is an exciting new online tool that provides essential satellite and mapping information for farm producers across Canada.
Today, Canadian farm producers are using field and production data more than ever to manage their operations. Different types of data allow producers to keep a close eye on crop and pasture production, track weather and soil conditions on their land, meet traceability requirements and a whole lot more. While a lot of remote sensing data are publicly available through NASA and ESA, it can be difficult to access without special software and training. By using AgSat, farm producers can access these essential data sources to optimize their agricultural production decisions, and pinpoint areas on their land that may need more attention.
Currently low resolution static basemaps are used to help users locate areas of interest and use map tools to order satellite imagery. The winning solution to this challenge would be used to replace or supplement existing basemaps within the AgSat application.
If sufficiently frequent update cycles can be achieved (e.g. sub-monthly), the basemap could also be used as an indicator of potential changes on the managed land by agricultural producers or watershed managers, in order to guide acquisition of higher-resolution (paid) satellite imagery.
How do I compete?
Follow the next few steps:
Visit the challenge dashboard to sign up: https://ai.science/la/cohort9-202105 (further details including FAQ available here.)
Set up an Account or Sign In → Click “Register Now”
Find teammates through your network or through the Aggregate Intellect Slack Workspace (cohort-9-support channel)
Register your team under the “Form Your Team” tab
Individually take 5 minutes to complete the self-assessment
Collaborate with your teammates to complete a joint proposal (download the template & watch the AI product videos as a guide)
Submit your proposal
If your proposal is accepted, you’ll get to join the challenge for free!
Details:
This challenge is hosted by Aggregate Intellect, an online marketplace for accelerated product development. This platform hosts a global community of industry practitioners and researchers gathered around topics related to research, engineering, and product development in AI, and some other emerging technologies.
Register on Aggregate Intellect
Aggregate Intellect uses the collective intelligence of the crowd along with an AI-based toolkit to enable developers build innovative products faster, and better. This enables startups with limited resources to experiment with various ideas rapidly, and validate the most promising ones for further development in a cost and effort efficient way. The platform allows developers to flex their technical muscles by working on challenging problems put forward by startups, while helping startups achieve more meaningful time to market for their newest product features.
You can join competitions/challenges like this, as well as other activities like discussion groups for free, and get access to a full suite of content offered by the platform (papers, videos, code snippets, etc.). Further details including FAQ available here.
If you have any questions, contact the organizers at challenges@ai.science
Important Dates and Resources
This product challenge officially begins on May 17th, 2021, but the latest date to submit a proposal is actually on May 14th, 2021! Once the challenge begins participants will have approximately one month to produce their solution to the challenge, with a final submission deadline on June 13th, 2021.
Last Chance to Submit Proposal: Friday, May 14th
Cohort Begins: Mon, May 17th
Final Submission Deadline: Sunday, June 13th
Watch a recording of the kickoff meeting below to learn more about the challenge: