Talk with an expert


View fullsize

Leveraging Farmwave’s image recognition capabilities, reduce the time to count wheat seedlings in the field by 25% while maintaining and accuracy of 75+%. In 2018, Dijon Céréales approached Farmwave. They needed a better way to count wheat and barley at early growth stages to increase accuracy in forecasting yield prediction. Using manual methods, accuracy decreased due to human error and fatigue. Farmwave’s image recognition and AI model capabilities were sought to provide a baseline capability which would feature high accuracy and consistency.


Dijon Céréales collected 75 relevant images, 50 of which were presented to Farmwave for training an AI model for this project. Dijon Céréales retained 25 images in reserve for testing purposes at the end of the project.

View fullsize


Despite the time and imagery limitations, Farmwave achieved a 70-75% accuracy rate via its counting capabilities compared to the hand-counted model.


  • Consistent imagery is quality is required to train accurate models.

  • A Standard Operating Procedure (SOP) is needed for training in-field specialists.

  • Higher accuracy than the current 70-75% can be obtained sufficient imagery, and more time