Release notes

We've integrated one of the leading map providers, HERE Maps, into our catalog search and job configuration screen. This integration brings a much nicer search experience, as well as a more feature-rich and aesthetically pleasing basemap!

The integration with HERE Maps brings much more advanced geocoding, which among other enhancements, provides better performance for non-Latin words. We can now also support searching based on coordinate formats, such as 45.314483, 30.714823 or 45°18'52.1" N 30°42'53.4 "E. You can search based on coordinates, and the platform will simply go ahead and select a small AOI around that point. You can then manipulate that AOI to fit your requirements. Of course, users can still upload their AOI in GeoJSON format and have their AOI automatically selected!

Search for data

Back in December, we launched our price estimation - one of our most hotly-requested features. Backed by data on the thousands of jobs that UP42 customers have run previously, our data science team built a machine-learning algorithm and a basic rule set algorithm that provides an estimated range of data, processing, and infrastructure costs.

This estimate is available both within the console or through the Python SDK, leveraging the command, workflow.estimate_job(input_parameters).

This estimate, along with the prices given during the data ordering process within the catalog search, is now used to ensure that you don't send your account spiraling into minus, even when you're running multiple jobs or ordering multiple data sets concurrently. As a result, you and your team can concentrate on analysis or data acquisition, running jobs, and ordering data with confidence, knowing that you won't end up with an account in minus.

Run a job

This block from our new partner and leader in forestry inventory management solutions, Tesera, generates GeoJSON polygons that follow major land surface features, including water bodies, rivers, vegetation cover, and types. It uses Sentinel-2 multispectral imagery at 10-meter resolution as an input, generating two sets of land line polygon segments that average 1 hectare and 10 hectares, respectively.

Specifically, this processing block uses the Sentinel-2 L2A Analytic (GeoTIFF) block from UP42 that provides bottom-of-atmosphere reflectance values that are analysis-ready. This increases the reliability and reduces errors in the results caused by atmospheric interference and other non-feature-related noise that can be present in uncorrected satellite imagery.

The polygon output can be used for multiple use cases. Firstly, analysis can delineate vegetation changes and disturbances, like fires or cutblocks, at much a larger scale faster and lower than airborne remote sensing. The land lines can also be used as a pre-processing step for classification use cases, such as land cover classification or forestry mapping.

Use the block

We're excited to launch this state-of-the-art field boundary detection algorithm, provided by Airbus and Agrimetrics. The FieldFinder processing block runs on SPOT imagery and accurately delineates agricultural fields, outputting a shapefile.

FieldFinder applied to SPOT imagery in the UK

The block has been rigorously trained and tested across many European countries and the entire state of Iowa, US, enabling consistent wide area coverage. Boasting a mean F1 value of 0.85 in the regions shown below, FieldFinder is a reliable model for detecting even small fields.

The model accuracy and use of SPOT imagery rather than Sentinel-2 differentiate the algorithm from other automated boundary detection solutions and ensure it meets the cartographic quality demands for users' from crop growers to developers building agricultural solutions.

If the FieldFinder blocks don't currently cover your required region, get in touch, and we'd be happy to arrange model training to fulfill your requirements.

FieldFinder Europe | FieldFinder Iowa

Our partner and a leader in vegetation management analytics, Spacept, has released a new and improved version of all three of their algorithms - Tree Detection and Height From Shadow, Tree Detection, and Shadow Detection.

The algorithm benefited from significant accuracy improvements in the Tree Detection and Shadow Detection algorithms, where the Shadow Detection block has been benchmarked against human annotators on 10,000+ Pléiades images. The Shadow Detection model showed a precision of 91%, recall of 94.5%, and an F1 score of 93%.

These developments make Spacept's capabilities a compelling replacement or complementary technology for existing vegetation management processes for railway operators and utility providers. You can find out more about how satellite-based vegetation management solutions compare to existing LiDAR and new UAV technologies in our recent blog post.

Check out the blocks