Photometric stereo training dataset with annotated leaf masks

Automated leaf segmentation is a challenging area in computer vision. Recent advances in machine learning approaches allowed to achieve better results than traditional image processing techniques (Scharr et al., 2016; Mengye and Zemel, 2016); however, training such systems often require large annotated datasets (Giuffrida et al., 2017). To contribute with annotated datasets and help to overcome this bottleneck in plant phenotyping research, here we provide a novel photometric stereo (PS) dataset with annotated leaf masks. This dataset forms part of work done in the BBSRC Tools and Resources Development project BB/N02334X/1.