[Proposal-ML] Global Wheat Spike Detection and Quantitative Analysis Based on Image Recognition

Xun Wang, Tianhai Liang, Gu Zhang

Keywords: Wheat spike detection; Object detection; Adversarial training; Diffusion model; YOLO-v11

TL;DR: We proposed a framework leveraging adversarial training between a diffusion model and YOLO-v11 to enhance the accuracy of global wheat spike detection and counting.

Abstract: Wheat is one of the most important staple crops worldwide. Rapid and accurate detection of wheat spikes from outdoor field images is essential for farmers to implement efficient field management and enhance the quality of wheat cultivation. However, factors like planting environment and image quality complicate achieving robust and precise detection using visual models. Therefore, we propose a framework that employs adversarial training with diffusion models and YOLO-v11 visual models to address this challenge.

Submission Number: 39