Plant traits prediction for ecosystem modeling: Combined model based on images and auxiliary data

Abstract

The study presents a combined machine learning model for predicting plant traits. The model combines a pre-trained convolutional neural network (CNN) with auxiliary variables represented as comma- separated values (CSV). The aim of the study is to improve the accuracy of plant trait prediction using both visual information and contextual data.

The model is trained on a dataset consisting of 9,147 plant images and the corresponding 167 auxiliary variables. The images are preprocessed using three pre-trained convolutional neural network architectures (InceptionV3, ResNet, VGG19). The supplementary data are combined using a concatenate layer after directing the outputs of the above CNNs. The model is optimized using the Adam algorithm and evaluated with appropriate metrics.

The results show the superiority of the combined model over the baseline CNN in predicting plant traits, confirming the effectiveness of using supplementary data to improve performance. The study demonstrates the potential of hybrid machine learning models to analyze plant data from users around the world.