Annotated patches of whole oilseed rape (Brassica napus) plant images created using the MapReader pipeline

<p><strong>Background and Dataset Creation:</strong></p> <p>Patches derived from whole images of oilseed rape (<em>Brassica napus</em>) plants from the <a href="https://research.aber.ac.uk/en/datasets/collection-of-side-view-and-top-view-rgb-images-of-brassica-napus">'Collection of side view and top view RGB images of Brassica napus from a large scale, high throughput experiment'</a> dataset and their associated annotations, which were used to train, validate and test patch classification models as described in the following paper:</p> <p>Corcoran, E., Hosseini, K., Siles, L., Kurup, S., and Ahnert, S. 2024. 'Automated dynamic phenotyping of whole oilseed rape (Brassica napus) plants from images collected under controlled conditions', Frontiers in Plant Science (under review).&nbsp;</p> <p>Patches were created and annotated using the <a href="https://github.com/Living-with-machines/MapReader">MapReader</a> pipeline. Please see:&nbsp;</p> <ul> <li>Kasra Hosseini, Daniel C. S. Wilson, Kaspar Beelen, and Katherine McDonough. 2022. MapReader: a computer vision pipeline for the semantic exploration of maps at scale. In Proceedings of the 6th ACM SIGSPATIAL International Workshop on Geospatial Humanities (GeoHumanities '22). Association for Computing Machinery, New York, NY, USA, 8&ndash;19.&nbsp;<a href="https://doi.org/10.1145/3557919.3565812" rel="nofollow">https://doi.org/10.1145/3557919.3565812</a></li> <li>Kasra Hosseini, Rosie Wood, Andy Smith, Katie McDonough, Daniel C.S. Wilson, Christina Last, Kalle Westerling, and Evangeline Mae Corcoran. &ldquo;Living-with-machines/mapreader: End of Lwm&rdquo;. Zenodo, July 27, 2023.&nbsp;<a href="https://doi.org/10.5281/zenodo.8189653" rel="nofollow">https://doi.org/10.5281/zenodo.8189653</a>.</li> </ul> <p><strong>File structure:</strong></p> <p><em><strong>Annotations</strong></em></p> <p>The <strong>'annotations_six_label_sv_5.zip'</strong> folder contains annotations for the entire patch dataset in .csv format, these files have two columns <strong>'image_id'</strong>, <strong>'label'</strong> in which:</p> <ul> <li><strong>'image_id'</strong> = the path to each image patch</li> <li><strong>'label'</strong> = the label assigned to each patch by the annotator indicated which part of the plant the patch primarily contained, or if it was part of the background. Labels: <strong>'0'</strong> = non-plant background, <strong>'1'</strong> = open flower, <strong>'2'</strong> = flower bud, <strong>'3'</strong> = leaf, <strong>'4'</strong> = greed pod containing seed, <strong>'5'</strong> = branch.&nbsp;</li> </ul> <p><em><strong>Patches</strong></em></p> <p>The 'b_napus_patch_data.zip' folder contains all patches in csv format. Each file is named in a consistent format e.g. "patch-1580-330-1590-340-#2018-07-06_00_VIS_sv_000-0-0-0.png#.PNG" where '1580-330-1590-340' are the x and y coordinates of the patch boundary and '#2018-07-06_00_VIS_sv_000-0-0-0.png#' indicates the image in the <a href="https://research.aber.ac.uk/en/datasets/collection-of-side-view-and-top-view-rgb-images-of-brassica-napus">'Collection of side view and top view RGB images of Brassica napus from a large scale, high throughput experiment'</a> from which the patch was derived.&nbsp;</p>