Why we can’t upgrade existing systems to Phena 2.0 (yet)

We created Phena 2.0 to optimize the quality of our data. But, for now, we cannot upgrade Phena on most of our existing clients’ systems. In this blog we will explain why we cannot upgrade the systems and what upgrade options you still have if you really, really want to upgrade Phena.

Keep in mind that HortControl will still be updated, and you’ll benefit from all other features we release, such as BrAPI.

Phena

We call our processing chain – from raw sensor data up to plant parameters – Phena. We’ve been optimizing this processing chain ever since we developed the first PlantEye sensor 10 years ago. Phena merges, triangulates and optimizes 3D files, adds additional calculated spectral information (like NDVI values) then calculates the plant parameters.

We’ve drastically restructured Phena to be more flexible and add even more information to our 3D files, allowing us to improve now, and making it easier to develop in the future.

Why can’t we upgrade existing systems?

In fact, we can and want to upgrade all our systems to Phena 2.0, but there are two good reasons for our clients not to do so now.

1. Loss of data / Old data on a mirrored system
2. Models need to be re-trained / R&D pipelines need to be adjusted

Loss of data / Old data on a mirrored system

With the development of Phena 2.0, we also made a major change to the HortControl database structure. Changing the database allows us to apply to the data standards of MIAPPE. It gives us more flexibility to add new PlantEye parameters as well as data from other sensors like DroughtSpotter.

However, right now, we have no options to transfer data from the old database to the new one. Upgrading now would mean losing all your historic data or having to create a mirrored version of your old HortControl to retain access. Unfortunately, in the latter scenario, we would not be able to migrate your old data when we do develop a data migration tool.

Models need to be re-trained / Development pipelines need to be adjusted

Phena 2.0 and Phena 1.0 have different outputs for processed 3D files, so the values of the plant parameters calculated from these processed 3D files are also different.

Many (machine learning) models depend on training data from both PlantEye and the target value (the value the model must predict from the PlantEye data). The models are trained on data from Phena 1.0, which means your model will have a different output value when making predictions based on Phena 2.0 values. Models must be re-trained and re-validated to correct this, which can consume a lot of time and resources. These models have been working perfectly well for many of our clients, and we understand there is no need to upgrade or invest additional resources.

In a case similar to the models, many of our industry clients created or adjusted their R&D pipelines based on our data. Decision-making algorithms and threshold limits are based on Phena 1.0 data, and the data from Phena 2.0 could easily lead to different decisions/outcomes. Also, upgrading and adjusting a working development pipeline is not really in our clients’ interest.

Nevertheless, in some cases, the increase in parameter quality and the number of parameters might also significantly influence the quality of machine learning models and product development pipelines. So, we’ve put together some upgrade options for you to consider.

The upgrade options you do have

Depending on your situation, we have a few options for those who really want them. Get in touch with your contact person to discuss them in detail.

Full clean install

A complete clean installation with a new empty database ensures a smooth-running system. Consider this option when:

• The data in your current HortControl database is of no value anymore.
• Your system was just installed, and you only ran a few trial experiments.
• You did not create any machine learning or R&D pipelines that rely on Phena 1.0 data.

In these situations, waiting for data migration or running a mirrored HortControl installation will make no difference.

Full clean install, with a mirror version of the old HortControl installation

We are talking to our clients to define the need for a data migration tool. If this is the case, we will plan the data migration tool for our next HortControl release. Consider this option when:

• You want to keep access to the data in your current HortControl database for reference
• You did not create any machine learning models or R&D pipelines depending on Phena 1.0 data

Wait for the migration tool 

We are talking to our clients to define the need for a data migration tool. If this is the case, we will plan the data migration tool for our next HortControl release. Consider this option when:

• The data in your current HortControl is vital, and you need to retain unrestricted access to it.
• You did not create any machine learning models or R&D pipelines that rely on Phena 1.0 data

Do not upgrade (ever)

Phena 1.0 already delivers high-quality data and, if the following situations are true for you, there is no reason to upgrade at all

• Your (machine learning) models are good and re-training the models would be costly.
• Your R&D pipelines are running smoothly and adjusting the pipelines would be costly
• And there is no need to improve the models or pipelines with the enhanced 3D files or new parameters