Sharing the Right Data Right: A Symbiosis with Machine Learning

Abstract

In 2014 plant phenotyping research was not benefiting from the machine learning revolution as appropriate data were lacking. We report the success of the first open dataset in image-based plant phenotyping suitable for machine learning, fuelling a true interdisciplinary symbiosis, increased awareness and steep performance improvements in key phenotyping tasks. Advancing plant phenotyping by sharing 'problems' Appropriate training and testing data are at the heart of computer vision (CV) and machine learning (ML) research as means for developing and evaluating novel approaches. However, in 2014, appropriate data for image-based phenotyping problems were lacking. Thus, plant phenotyping was not benefiting from this data-driven revolution [1] and CV/ML researchers were largely unaware of phenotyping applications. To address these limitations, we opened a collection of plant data for several phenotyping tasks including two 'hot' CV/ML problems: leaf segmentation a 'multi-instance segmentation' problem and leaf counting an 'object counting'

Crop Phenomics and High-Throughput Phenotyping: Past Decades, Current Challenges, and Future Perspectives

“…(3) data sharing: sharing the right data and the right questions to attract more computer specialists to help solve problems for free, even if the solutions are not perfect (Tsaftaris and Scharr, 2018).…”

Leaf Counting Without Annotations Using Adversarial Unsupervised Domain Adaptation

“…As of now, deep neural networks are the state-of-theart of machine learning models, as they have been demonstrated to work well for plant phenotyping problems as well [37], such as leaf segmentation [26,27,41,44] and leaf counting [1,7,10,16] show extraordinary performance. Although deep learning has been demonstrated to perform well on the leaf counting, still lacks of generalisation on unseen datasets.…”

Global Wheat Head Detection 2021: An Improved Dataset for Benchmarking Wheat Head Detection Methods

“…Similarly, the importance of getting plant-or crop-specific datasets is recognized within the plant phenotyping community ( [4][5][6][7][8][9][10], p. 2, [11][12][13]). These datasets allow benchmarking the algorithm performances used to estimate phenotyping traits while encouraging computer vision experts to further improvement ( [10], p. 2, [14][15][16][17]). The emergence of affordable RGB cameras and platforms, including UAVs and smartphones, makes in-field image acquisition easily accessible.…”