Research Article | Open Access

Volume 2023 |Article ID 0073 | https://doi.org/10.34133/plantphenomics.0073

Deep Learning Enables Instant and Versatile Estimation of Rice Yield Using Ground-Based RGB Images

Yu Tanaka ,^1,2,18 Tomoya Watanabe,³ Keisuke Katsura,⁴ Yasuhiro Tsujimoto,⁵ Toshiyuki Takai,⁵ Takashi Sonam Tashi Tanaka,^6,7 Kensuke Kawamura,⁵ Hiroki Saito,⁸ Koki Homma,⁹ Salifou Goube Mairoua,¹⁰ Kokou Ahouanton,¹⁰ Ali Ibrahim,¹¹ Kalimuthu Senthilkumar,¹² Vimal Kumar Semwal,¹³ Eduardo Jose Graterol Matute,¹⁴ Edgar Corredor,¹⁴ Raafat El-Namaky,¹⁵ Norvie Manigbas,¹⁶ Eduardo Jimmy P. Quilang,¹⁶ Yu Iwahashi,¹ Kota Nakajima,¹ Eisuke Takeuchi,¹ Kazuki Saito ^5,10,17,18

¹Graduate School of Agriculture, Kyoto University, Kitashirakawa Oiwake-chou, Sakyo-ku, Kyoto 606-8502, Japan
²Graduate School of Environmental, Life, Natural Science and Technology, Okayama University, 1-1-1, Tsushima Naka, Okayama 700-8530, Japan
³Graduate School of Mathematics, Kyushu University, 744, Motooka, Fukuoka Shi Nishi Ku, Fukuoka 819-0395, Japan
⁴Graduate School of Agriculture, Tokyo University of Agriculture and Technology, 3-5-8 Saiwaicho, Fuchu, Tokyo 183-8509, Japan
⁵Japan International Research Center for Agricultural Sciences, 1-1 Ohwashi, Tsukuba, Ibaraki 305-8686, Japan
⁶Faculty of Applied Biological Sciences, Gifu University, 1-1 Yanagido, Gifu 501-1193, Japan
⁷Artificial Intelligence Advanced Research Center, Gifu University, 1-1 Yanagido, Gifu 501-1193, Japan.
⁸Tropical Agriculture Research Front, Japan International Research Center for Agricultural Sciences, 1091-1 Maezato, Ishigaki, Okinawa 907-0002, Japan
⁹Graduate School of Agricultural Science, Tohoku University, Aramaki Aza-Aoba, Aoba, Sendai, Miyagi 980-8572, Japan.
¹⁰Africa Rice Center (AfricaRice), 01 BP 2551 Bouaké, Côte d'Ivoire
¹¹Africa Rice Center (AfricaRice), Regional Station for the Sahel, B.P. 96, Saint-Louis, Senegal
¹²Africa Rice Center (AfricaRice), P.O. Box 1690, Ampandrianomby, Antananarivo, Madagascar
¹³Africa Rice Center (AfricaRice), Nigeria Station, c/o IITA, PMB 5320, Ibadan, Nigeria
¹⁴Latin American Fund for Irrigated Rice - The Alliance of Bioversity International and CIAT, Km 17 Recta Cali-Palmira, C.P. 763537, A.A. 6713, Cali, Colombia
¹⁵Rice Research and Training Center, Field Crops Research Institute, ARC, Giza, Egypt
¹⁶Philippine Rice Research Institute (PhilRice), Maligaya, Science City of Muñoz, 3119 Nueva Ecija, Philippines
¹⁷International Rice Research Institute (IRRI), DAPO Box 7777, Metro Manila 1301, Philippines
¹⁸These authors contributed equally to this work.

Abstract

Rice (Oryza sativa L.) is one of the most important cereals, which provides 20% of the world’s food energy. However, its productivity is poorly assessed especially in the global South. Here, we provide a first study to perform a deep-learning-based approach for instantaneously estimating rice yield using red-green-blue images. During ripening stage and at harvest, over 22,000 digital images were captured vertically downward over the rice canopy from a distance of 0.8 to 0.9 m at 4,820 harvesting plots having the yield of 0.1 to 16.1 t·ha−1 across 6 countries in Africa and Japan. A convolutional neural network applied to these data at harvest predicted 68% variation in yield with a relative root mean square error of 0.22. The developed model successfully detected genotypic difference and impact of agronomic interventions on yield in the independent dataset. The model also demonstrated robustness against the images acquired at different shooting angles up to 30° from right angle, diverse light environments, and shooting date during late ripening stage. Even when the resolution of images was reduced (from 0.2 to 3.2 cm·pixel−1 of ground sampling distance), the model could predict 57% variation in yield, implying that this approach can be scaled by the use of unmanned aerial vehicles. Our work offers low-cost, hands-on, and rapid approach for high-throughput phenotyping and can lead to impact assessment of productivity-enhancing interventions, detection of fields where these are needed to sustainably increase crop production, and yield forecast at several weeks before harvesting.