As harvest season comes to a close, and the year is winding down, I wanted to provide the first Farmwave Harvest Vision 2020 Season Report. I hope everyone is well, staying safe, and looking forward to 2021.
It was a very exciting year for Farmwave. Bringing the Harvest Vision System to market had many challenges for this small team. As data scientists and software engineers, we had to become hardware engineers and fabricate everything used to show the AI capabilities. Our prototype from 2019 morphed into something we are proud of calling an MVP. In a COVID-19 environment we had thought about not doing it this year. Time, funding, parts, shipping delays... were all against us. This blind-sided everyone this year.
But, none of this was going to stop the farmers from having to harvest... so it wasn't going to stop us. As we said in the Marines - adapt and overcome.
We limited the number of deployments this year to control support. We were pleasantly surprised and stressed at the same time to hit this number in less than 4 weeks, (pre-orders for 2021 are already coming in). We did this limited number this year because we knew we would have some hardware flaws; and we did. Faulty power buttons, 3D printed chassis breaking during shipping, and a wi-fi issue we never anticipated from harmonic feedback and static electricity build-up in the back of the machine while it's threshing, (already solved for 2021).
But these were all hardware issues - an area not what Farmwave focuses on. All of these were fixable and we made all attempts to provide the promised high-touch service to compliment our high-tech product. In some ways I feel as though we failed... upward. We learned a lot and I could not be more proud of my team who built something in less than one year from ideation, to prototype , to field trials, re-design, MVP, and go to market... 10 months, 10 people, and lots of coffee.
But the AI did exactly what it's supposed to do. Because THIS is where the value is.
The attached report shows the results of this effort; almost $1M in retained crop in only 10 weeks on only 15 farms.
Seven years of AI innovation, and our two awarded patents, (#1 using machine-learning to identify a crop infestation and recommend a treatment and pesticide application and #2 using machine-learning to predict crop yields based on photos taken of the crop), millions of images and now our internal tooling process which we're calling the <50 rule, (we've automated annotation, learning, and accuracy with 50 or less images for any problem classifier), has brought us to the value we see in this report. All of this has accelerated the efforts for Sprayer Vision in the spring of 2021 as well as another redesign for 2021 Harvest Vision which speeds things up and increases functionality including new features like grain quality, measurement, and foreign material classification... in real-time while rolling through the field.
Thanks for your time and I hope you found some value in this report.