Picterra – NOAH19 London

Picterra – NOAH19 London


Hello everyone, thank you for still being here. It’s nice So 6,000 This is the number of satellites currently taking picture of the planet all these guys They are collecting 100 terabyte of data every day It means that in two days they have acquire as much data as the all Wikipedia in 40 languages So when it comes to satellite imagery we often have in mind Now inmate really taken from very far which they are obviously but with very close resolution actually, actually this is to picture of the same place in Antwerp in Belgium and today satellite imagery Looks like more the image on the top. So it’s 30 centimeter resolution and you can see that we can See actually a lot of details like I can tell or we could tell Which oil tank is full which one is empty how many ships? There is there a lot of activities on the pot? another example, which is a bit Mossad, this is a picture of The freight station in Brazil. So this is what their frustration looks like at saachi’s centimeter resolution so now imagine Imagine a search engine where that you can just you know, Curie or that you can just tap into this huge library of imagery and get information in real time like How many? Houses have been Billy Lin Jiang how many containers have been landed today in Antwerp all this kind of information? Well, the only issue with that is that we need a software which is able to recognize this type of feature and it’s a bit like a kid this software needs to see a lot of example Before being actually able to recognize this feature We said that in average a software I I dropped the word an algorithm a dip learning model Need between like ten to fifteen thousand example for each feature beef being able to recognize it Yeah, that’s that’s a pain So well, we did that picked our we created a new way to Train machine learning models that doesn’t need thousands of example, but that only need like five example the idea is that the user will show example to the platform and based on this example the platform will learn how to detect and to localize this feature and we have make it easy and accessible to everyone So basically thanks to pick tera anyone can now map any object on earth all in just a few clicks without being a data scientist or Developer or just yeah just made for normal people What we’ve done is that we have commoditized Machine learning and our users while they’re using picture. They are creating strategic assets for the company. Meaning they are creating like Deep learning models or that we call detectors Instead of having a bunch of data scientist and Joe spatial program. They just built like deep learning model in few clicks and For a fraction of a cost of course. So what is good, which picture is the network effect behind it means more user we have The more example the create the more this example make the algorithm better and then ultimately this algorithm Make the product better which attract new user and so on This is of course a scalable SAS platform but on the top of that this is quite defensible because the more you use it the more It’s the better it became actually So unless in less than six months The platform is is getting 10 times faster and two times better meaning more accurate so what I want you to to live with is that On pizza, we don’t sell a product. We don’t sell Subscription we sell basically the capacity to tap into a non-touch reservoir of info of geospatial information That anyone meaning researcher scientist developers even investors Could use to build application new application So when you have this in mind You realize that the market of pizza is not made of this little niche of 80 billion, which is the geospatial industry But the mass market could be in terms of economic impact value could be around 1000 billion dollars So thank you very much. I’m derek from peter. I we base in the son

You May Also Like

About the Author: Oren Garnes

Leave a Reply

Your email address will not be published. Required fields are marked *