Data for WeedScan manuscript

Computer vision model, scripts, and test images for benchmarking test described in the manuscript "WeedScan, a weed reporting system for Australia using an image classification model for identification".\nLineage: The WeedScan model was trained by Ron Li (Data61) on an image dataset comprising ca. 195,000 images of 488 taxa of weeds and Australian native relatives using MMClassification on Bracewell. The Seek model was obtained from the v2.9.1 release on its Github. The test image library was collated by Alexander Schmidt-Lebuhn mostly from his own photographs but also donations from NSW DPI and John Hosking.

本数据集包含用于复现论文《WeedScan：一款基于图像分类模型实现杂草识别的澳大利亚杂草上报系统》中所述基准测试的计算机视觉模型、脚本与测试图像。
模型谱系：WeedScan模型由Data61的Ron Li基于约19.5万张涵盖488类杂草及其澳大利亚本土近缘类群的图像数据集训练完成，训练过程依托Bracewell平台与MMClassification框架。
Seek模型源自其GitHub仓库发布的v2.9.1版本。
测试图像库由Alexander Schmidt-Lebuhn整理，素材主要来源于其个人拍摄的照片，同时也包含来自NSW DPI与John Hosking捐赠的图像。