Wheat Head Detection using Deep, Semi-Supervised and Ensemble Learning

In this paper, we propose an object detection methodology applied to Global Wheat Head Detection (GWHD) Dataset. We have been through two major architectures of which are Faster R-CNN, and EfficientDet, in order design a novel robust wheat head model. emphasize on optimizing the performance our proposed final architectures. Furthermore, extensive exploratory data analysis, cleaning, splitting adapted best augmentation techniques context. use semi supervised learning, precisely pseudo-labeling, boost previous models detection. Moreover, put much effort ensemble learning including test time augmentation, multi-scale bootstrap aggregating achieve higher performance. Finally, weighted boxes fusion as post processing technique optimize results. Our solution has submitted solve research challenge launched GWHD Dataset was led by nine institutes from seven countries. method ranked within top 6% above-mentioned challenge.