New technologies have been changing every industry and agriculture is no exception. Africa Agri Tech (AAT) unites the southern African agricultural communities to discuss the adoption of the latest technologies, innovations, and advances in AgTech.
The video below provides insight on how Google as already achieved this on existing projects
One of the guest speakers at their event this week is data engineer and pioneer within the DotModus team Hardus Swanepoel. He will be breaking down the use of Machine Learning in agriculture and how Google Cloud has been leading the race in providing famers and developers with the tools needed to identify and monitor the environmental and ecological changes affecting crops.
Hardus will be giving an overview of Google Cloud Platform and unpacking 4 major tools used to process data with machine learning models. These tools are Cloud Storage, Cloud Vision API, Cloud ML and Cloud Composer. When used in conjunction with one another, engineers are able to save farmers time and alert them to information in their data that may have otherwise gone unnoticed.
As seen in the linked video, Google Cloud was used to facilitate the whole process. Initially, cameras and drones are used to capture high resolution images that are then uploaded to Google Cloud Storage. In this use case there were many data sources including IoT devices, Cloud Storage enables engineers to pull information from multiple data sources at a time to identify the core root of problems in real-time.
Cloud Vision API provides the eyes of the operation. In the video , thousands of hi-res photographs are provided to Cloud Vision API for it to analyse and understand. The system takes this imagery and converts it to information for Cloud ML which in turn decides what to do with it. Cloud ML drives the decisions made by Vision API. It has been trained to identify and pin-point features that may indicate an unhealthy plant. If Vision API picks up one of these features, it flags it in the database. This has been used extensively in the agricultural space but can be used in just about any industry to automate processes and solve key business issues.
Once data has been gathered by Vision API and analysed by Cloud ML it is stored in Cloud BigTable (an analytical tool capable of ingesting terabytes of data in seconds). BigQuery is used to provide a mechanism for searching, filtering and exploring this stored information. This can be taken a step further (as seen in the video) by overlapping these results into Google Maps so they can be visualised over the crops.
Finally, automation is handled in Google Cloud Composer that schedules when tasks are required to happen. Automation can be carried across various verticals and enables the user to plan drone routes for instance. It also allows the user to schedule when they will be receiving data. For instance, data can be ingested every minute to ensure a live stream of data or, alternatively, this can be passed in batches for use in a daily/weekly report.
Automating this process of image collection from hundreds of hectares of farmland is enabling farmers to take action earlier and, in the use case above, save almost 30% of their crop.