Picterra had a blast at Applied Machine Learning Days at EPFL. The last couple of days spent at the SwissTech Convention Center were full of exciting presentations, workshops, pitches, and getting to know machine learning professionals and enthusiasts from all over the world. An evening with keynote speakers such as whistleblower Edward Snowden made this edition especially unique! And all of that was at less than a kilometer from our office.

We had an opportunity to present the Picterra platform in front of a knowledgeable audience and to show some real-life examples of custom detectors. Our team members explained that with Picterra, it’s possible to train a model and detect various objects such as buildings, animals, crops, roofs, solar panels, road cracks or you name it – whatever users need. It’s easy to do it on the platform, in just a few clicks. On top of that, no costly training is necessary.

Applied Machine Learning Days at EPFL 2020

How did we do it?

• On Monday, 27 January 2020, our CTO, Dr. Frank de Morsier, gave a talk on the track “AI & Space.” He shared insights about the application of machine learning for mapping in Earth observation imagery and how Pictera democratizes access to Earth observation with machine learning. For more information, visit the AMLD’s website.

• On Tuesday, 28 January 2020, Frank de Morsier presented Picterra in the “AI & Industry” track where he shared about Picterra’s vision and latest achievements with large-scale detections exploiting our unique low-shot learning technology. This enables to train models with as low as a few dozen of training example annotations instead of the typical thousands required with traditional deep learning models.

• On Wednesday, 29 January, our Lead Software & Machine Learning Engineer, Julien Rebetez, presented Picterra during the startup pitches session. He shared a story from a user who was able to detect slurry tanks near 34’000 farms spread over the whole country of Denmark.

• Our team was present at 2 booths, ready to engage in the discussion about different applications of machine learning in Earth observation and leveraging our platform for fast prototyping and deployment in the production of detectors.

The previous edition has attracted over 4 thousand attendees, and although the organizers haven’t shared the statistics from 2020 yet, we expect their number to be much higher this year.

We can’t wait for the 2021 edition!