Plant Species Identification from Leaf Images Using Deep Learning Models (CNN-LSTM Architecture)

Abstract

Species knowledge is important for biodiversity conservation. Identification of plants by conventional approach is complex, time consuming, and frustrating tor non-experts due to the use of botanical terms. This is a challenge for learners interested in acquiring species knowledge. Recently, an interest has surfaced in automating the process of species identification. The combined availability and ubiquity of relevant technologies, such as digital cameras and mobile devices, advanced techniques in image processing and pattern recognition makes the idea of automated species identification become real. This paper elucidates development of convolutional neural network models to perform plant species identification using simple leaves images of plants, through deep learning methodologies. Training of the models was performed by using an open database of 100 plant species images, containing 64 different element vectors of plants in a set of 100 distinct classes of plant species. Several state-of the- art model architectures were trained, with the proposed model attaining performance of 95.06% success rate in identifying the corresponding plant species. The significant success rate makes the model very useful identifier or/and advisory tool. The approach could be further expanded to support an integrated plant species identification system to operate in real ecosystem services.