Sign up for our daily Newsletter and stay up to date with all the latest news!

Subscribe I am already a subscriber

You are using software which is blocking our advertisements (adblocker).

As we provide the news for free, we are relying on revenues from our banners. So please disable your adblocker and reload the page to continue using this site.
Thanks!

Click here for a guide on disabling your adblocker.

Sign up for our daily Newsletter and stay up to date with all the latest news!

Subscribe I am already a subscriber

New version Plantalyzer Vision software promises over 99% accuracy

HortiKey said it improved the Plantalyzer Vision software capabilities. The results are higher accuracy and higher consistency in tomato detection, according to the company. The new release comes with additional data: a truss counting feature. All currently operational systems will be updated with the newest Plantalyzer Vision software and will be replaced by a new computer.

Deep learning algorithm for tomato detection
After the first release in 2019, the development team started to work on the next-generation Plantalyzer Vision software. They upgraded the software with a deep learning algorithm for tomato detection. The challenge was to have the software and hardware fast enough to run the right cycle times. The result is an excellent solution with high capabilities in image analyses. While deployed, the system is able to analyze a huge amount of images, and as a result, results are available immediately after deployment.

99% plus accuracy
Results of the Plantalyzer Vision software are input for the crop estimation. Therefore significant validation is done to interpret the data. The outcome of the validation of the tomato detection in TOVs is an incredible accuracy of 99,7%. Verification on a second dataset of another variety TOVs from a different greenhouse confirms a 99% plus accuracy. The new Plantalyzer Vision software gives a very excellent basis for qualitative results for yield prediction!

Extension with truss counting
The deep learning algorithm for tomato detection also gives new possibilities. A first outcome is the extension of the software with a truss counting feature. It improves the end results because now the trusses are already counted on the robot instead of calculated in the dashboard.

Operational systems will be updated
The deep learning algorithm and truss counting are integrated into the Plantalyzer Vision software. While the new release requires more graphic processing performances, a new computer will be installed on all operational systems. All systems will be updated with the new Plantalyzer Vision software release.

Data analytics
Plantalyzers are deployed in the market for the third year now. This year all source data, including all raw images, of a complete cultivation season have been saved. This unique dataset enables new opportunities to the development team to work on data analytics and simulation for new insights independent of operational use.  

Ready for other tomato varieties
Besides the improved performance, the new Plantalyzer Vision software is ready to be more applicable to other tomato varieties. Validation of the Vision software for other tomato varieties is currently done in practice at several greenhouse growers in the market. Andreas Hofland, General Manager HortiKey: ‘The new Plantalyzer Vision software is a major leap for Plantalyzer performances, and we are very happy to release it to the market.’

For more information:
HortiKey
Burgermeester Crezeelaan 42A
2678KZ De Lier, Netherlands
T +31 (0)174517700
hortikey.nl

Publication date: