Wheat Heads Detection using Deep Learning Algorithms

Abstract

This paper aims to analyze the use of object detection in the field of wheat phenotyping and the components used inside it. Recently computer vision has started to play a pivotal role in agriculture, so the special emphasis has been given to the state-of-the-art object detection models to compare them against one another. The comparison of each algorithm is performed based on performance on a dataset known as GWHD and its Box AP and APS values on the benchmark dataset. The most efficient networks for object detection, comprising of a single-stage detector known as YOLO and a two-stage detector known as Faster R-CNN, have been studied, in general, and compared on many fronts to get an overall and comprehensive comparison.