Monitoring the characteristics of the wheat heads regularly will assist the breeders to grow better varieties and the farmers with better management decisions. However, this task is still done manually which is a time consuming and can be erroneous in some scenarios. When the popular deep learning based object detectors (Faster-RCNN, Yolo) are applied on 2D wheat head images, they tend to miss detecting the wheat heads that are influenced by motion blur due to wind and this impacts the accuracy of yield prediction and is not desirable. The students will be using baseline pytorch code (FasterRCNN) to study the influence of appropriate sharpening techniques to improve the wheat head detection accuracy. 

Interested students can contact me at kseemakurthy@lincoln.ac.uk