Auto-Tagging Soybean Threat Flights
In our last sprint, we have made changes in the product logic to only show the defoliation score when we detect it in the field and not present the insects chewing damage. This focus helps the users more quickly identify and process the main value-added pieces of information. A nice side benefit of this change in the product logic has been that we were able to auto-tag 30% of the tagging load (on average). This means more consistent and faster results for our customers as well as cost savings on manual tagging expenses.
Deep Canopy Segmentation
One of the main challenges of automatically tagging imagery is the clear distinction between the actual canopy and the rest of the image (i.e. ground/earth). We have made significant changes to our deep canopy segmentation model which will be released in this sprint to increase its accuracy and build the foundation for many of our future and current algorithms. A clearer distinction between the canopy and the rest of the picture will help us to more reliably tag not only soy defoliation but also many other threats that are currently being tagged manually.
Field Health Insight Report
In order to enable our growers and their advisors to get more out of our Field Health insight, we have changed the type of NDVI layer displayed in the report. Previously, the two layers from the different dates were drawn on different NDVI scales which made it impossible to compare. From this release onwards, we will show the NDVI layers using the same scale which enables growers and advisors to better compare the two dates, providing a much clearer understanding in regards to the growth of the biomass in the field.
Increased Security for John Deere Integration
We are constantly working to ensure our systems comply with the most advanced security standards. To that effect, we have upgraded our authentication protocol in our integration with John Deere to use OAuth 2.0. This protocol has become the industry standard for providing secure access to web APIs. It allows applications to access users' data without compromising security, and without the need to access user credentials in order to validate each API request, as was done on OAuth 1.0.
Taranis Artificial Intelligence: Improved Model for Diseases Detection
In order to further increase the consistency and speed of our SmartScout Insights, we have made improvements to our AI capabilities for disease detection. These improvements will enable us to more quickly and accurately detect a wide variety of diseases further improving the turnaround time for our Smart Scout Insights.
As more and more retailers and growers rely on our systems and data for critical business decisions, we are investing in making them more robust and reliable. In this release, we have completed several significant improvements to our disaster recovery systems including database and cloud storage backup. We have established a disaster recovery process of our various system components, allowing fast recovery with minimum system downtime.
Flight Plan Generation
In order to shorten our SLA for Insights generation, we are enabling our drone operators more flexibility in the flight plan generation process. From this release onwards, drone operators will be able to make changes to the flight plan such as setting take off points, update obstacles, changing flight parameters during the flight plan generation process simply by cancelling the current process and creating a new one. Previously the process had to be completed before making any changes which translated into more time spent in the field negatively impacting our SLA.
A Sneak Preview - coming soon
New Field Health Insight
We are moving to a new model for the field health insight. This model will present areas in the field with anomaly behavior related to the rest of the field. You will be able to see areas trending above or below field average when it comes to NDVI changes throughout the season. This new model will enable growers and their advisors alike to quickly and accurately identify those areas and further investigate the root cause(s) behind the anomalies. In order to dig deeper, growers / agronomists can decide to use the Taranis SmartScout Insights and/or manual scouting.
Support September Satellite Subscription
To support our growing customer base in the southern hemisphere, we will launch a subscription package to our premium satellite imagery with a start date of September. Previously, the annual subscription package was only available from January onwards which was optimized for our customer base in the northern hemisphere. The ability to choose an annual subscription package from September onwards is more suited for the seasons in the southern hemisphere. This is especially relevant since each annual subscription comes with access to historical images from the previous year. With this new change, both the current subscription as well as the image history are more aligned with the seasons in the southern hemisphere.
Early Flights Defoliation Severity
In the next sprint, we will be releasing a new defoliation severity algorithm based on the deep canopy segmentation that we released in this sprint. The new model will enable us to measure defoliation severity significantly better in the early soy stages (2nd flights) onwards.
We are also going to release our new M300 optimized disease/nutrition deficiency, detection model. This model will help us to improve our speed of delivery, increase consistency and reduce tagging costs by increasing our auto-tagging rate of these threats by ~20%.