AUTHOR=Fu Leiyang , Li Shaowen , Rao Yuan , Liang Jinxin , Teng Jie , He Quanling TITLE=A novel heuristic target-dependent neural architecture search method with small samples JOURNAL=Frontiers in Plant Science VOLUME=Volume 13 - 2022 YEAR=2022 URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2022.897883 DOI=10.3389/fpls.2022.897883 ISSN=1664-462X ABSTRACT=Crop classification has essential application value in germplasm resources and phenotype construction. Compared with traditional image processing methods, convolutional neural networks can identify features automatically. Nevertheless, varieties of crops and scenarios are complex, and it is not easy to obtain a universally applicable classification method. Manually designing neural network structures requires specialized knowledge and is time-consuming and labor-intensive. In contrast, neural architecture search methods can automatically search and create network architectures with parameter combinations when facing new objects. This paper collects eight kinds of rapeseed images to construct datasets (RSDS, rapeseed dataset). We propose a novel target-dependent neural architecture search method based on VGGNet (TD-NAS, target-dependent neural architecture search). Results show that the test accuracy of the model trained on small samples is not much different from that trained on large pieces. The generalization ability of the model generated by our method is not sensitive to the size of datasets, which makes it meaningful to quickly search model structures from small samples when facing new objects.