At Picterra we pride ourselves on allowing our users to quickly and easily build highly accurate machine-learning (ML) models to detect what’s important to them anywhere on Earth.
Our cloud-native platform allows them to achieve this without any coding or deep-learning knowledge, without costly infrastructure investment, and up to 95% faster than traditional methods.
As a cutting-edge company, we’re always looking for ways to enable our users to take advantage of the latest and greatest in geospatial and artificial intelligence (AI) technology and get even more out of their investments in satellite and drone imagery.
For this reason, we are excited to announce the integration of Meta AI’s Segment Anything Model (SAM) into Picterra making the production of ML models faster, easier, and more accurate than ever before.
Zero code, zero setup, works like magic
A crucial part of building an accurate custom detector, rather than relying on off-the-shelf or often out-of-date pre-trained models, is in the annotation process.
Annotating objects enables your Picterra model to learn about your specific use case within your specific geography, allowing for highly accurate feature extraction at scale.
With Picterra surprisingly few annotations allow you to build a model in minutes and hours rather than days and weeks. Even with our easy-to-use and efficient drawing tools, this can still take some time and consistency to ensure accurate results.
That’s where our new AI magic wand powered by SAM comes in!
Simply click on the new AI magic wand within the Picterra interface and you can start annotating trees, buildings, vehicles, cracks, or anything at all with just one click.
Cloud-native, no powerful GPU or costly IT infrastructure required
Once you have annotated a few examples of your land pattern or object of interest, model training can begin.
Since Picterra is an all-in-one cloud-native platform no installation of libraries or toolboxes is required and you don’t need to invest in a powerful GPU to carry out either this intelligent annotation or training.
Simply log in, import your imagery, one-click annotate, and train.
Other innovative Picterra features such as dataset recommendation will even suggest areas in which to continue one-click annotating to ensure you get the best results possible.
Create powerful custom detection models in minutes
With this exciting addition to the Picterra platform we are bringing ML production to all of our user’s fingertips.
Leverage the fastest annotation and training process in the industry to create powerful custom models for a wide range of industry applications.
Alongside this release, we’re also working on using SAM & similar models for various tasks across the platform, including increasing detector accuracy & providing better dataset recommendations.
We love to bring state-of-the-art ML research to our users in the most streamlined way possible. Don’t have a software engineering team to set up this kind of next-gen tech? We’ve got you covered! Our ML team will pick the best models & handle the infrastructure – all you need to do is bring your imagery & domain expertise, & see immediate results.
