Sourcing and Annotating Large Volumes of Agricultural Imagery for Precision Spraying

Disclaimer: Aya Data holds to the confidentiality agreements made with our clients. We shall not disclose any confidential information regarding clients and/or projects, pursuant to the individual agreements made. Consequently, client and project names and any identifiable information may be kept anonymous in our Case Studies.

Overview

Client X is a leading AgTech company focused on improving agricultural productivity and sustainability through cutting-edge technology solutions. Their aim was to have their in-house data science team develop a computer vision model that would detect and identify weeds.

Images of agricultural fields would be collected by drones and satellites. The CV model should be able to detect and identify weeds from the data collected, allowing farmers to optimize their weed management strategies, ultimately minimizing the impact weeds have on crop yield.

Data Acquisition, 

Data Annotation

Image segmentation

Industry

AgTech

Headquarters

London, UK

Company Size

250+

Challenge

The computer vision model would be trained by 100,000 annotated images of various crop and weed species. The Client sought the services of Aya Data to annotate the images.

During the project, it was discovered that:
a) The Client did not have all the required data for the annotation project.
b) The Client’s in-house data science team was overwhelmed by multiple tasks                 and falling behind schedule.

Contact us

Solutions

Aya Data acquired the additional data information necessary for the project by using our web scraping team to scrape the data from the internet. The additional data was curated and partially annotated using bounding boxes. For the next step, our data scientists created and trained a bespoke AI model to help with pre-labeling the images. 

The model performed well enough that the Client used it instead of their initial models.

Results

Due to Aya Data’s involvement, the Client developed a highly accurate CV model for weed detection on time and 21% under budget. Farmers were able to quickly identify and address weed infestations, which resulted in reduced crop loss and a more effective way of using herbicides.

Additionally, because the project was completed 21% under budget, the Client was able to reallocate the resources into R&D, enabling them to develop improved AgTech solutions for their customers.

The Client was satisfied that Aya Data’s services saved them resources and contributed to the successful completion of their weed detection project. The Client and Aya Data have established a long-term partnership for future projects relating to AgTech as a result of this project.