Machine-Learning Approach to Non-Destructive Biomass and Relative Growth Rate Estimation in Aeroponic Cultivation

We train and compare the performance of two machine learning methods, a multi-variate regression network ResNet-50-based neural network, to learn forecast plant biomass as well relative growth rate based onfrom short sequence temporal images from plants in aeroponic cultivation. The training dataset consists 57 taken different angles every hour during 5-day period. results show that top-down perspective produce better for while side are ResNet-50 network. In addition, using both cameras improves estimates but not those multivariatemulti-variatemultivariate regression. However, all were improved by cameras. found best produced model trained on top camera moving average filter resulting root mean square error 0.0466 g. 0.1767 g/(g·day).