Paddy Rice Imagery Dataset for Panicle Segmentation

Accurate panicle identification is a key step in rice-field phenotyping. Deep learning methods based on high-spatial-resolution images provide high-throughput and accurate solution of segmentation. Panicle segmentation tasks require costly annotations to train an robust deep model. However, few public datasets are available for rice-panicle We present semi-supervised model training process, which greatly assists the annotation refinement datasets. The learns features with limited localizes more positive samples datasets, without further interaction. After dataset refinement, number increased by 40.6%. In addition, we trained tested modern models show how beneficial both detection tasks. Results our comparison experiments can inspire others preparation selection.