Classification of Plant Species based Seedlings and Weedlings in Low Lightening Conditions using Deep Convolution Neural Network

Abstract

Smart farming techniques involve the use of plant identification and classification. Deep learning can be particularly useful for classifying low-light images because it can impulsively learn features from the data that can be relevant for classification. This is especially important in low light conditions where the image may be noisy or contain artefacts that are not relevant to the task. In the experiment, the plant seedlings and weedlings dataset consisting of low light images are subjected to a deep-learning model. Low-light images tend to have poor image quality due to the limited amount of available light. This results in a very low signal-to-noise ratio, making extracting beneficial information from the images extremely ambiguous. In the proposed work, a deep learning XceptionNet model is utilized to perform classification of plants using seedlings and weedlings that provides performance yielding an accuracy of 94.13% with 25 epochs.