Agricultural Informatics

7Deep Residual Neural Network for Plant Seedling Image Classification

Prachi Chauhan^1*, Hardwari Lal Mandoria1 and Alok Negi²

¹College of Technology, G.B. Pant University of Agriculture and Technology, Pantnagar, India²Department of Computer Science and Engineering, National Institute of Technology, Srinagar, India

Abstract

Efficient plant cultivation depends in large measure on weed control effectiveness. Weed conservation within the first six to eight weeks since planting is critical, because during this time weeds are competing aggressively with the crop for nutrients and water. In general, yield losses will range from 10 to 100% depending on the degree of weed control practiced. Yield losses are caused by weed interfering with growth and production of the crop. This explains why successful weed control is imperative. The first vital prerequisite to do successful control is accurate identification and classification of weeds. In this research we conduct a detailed experimental study on the ResNet to tackle the problem of yield losses. We used Plant Seedling dataset to train and test the system. Using ResNet (advanced convolution neural network) we classify the images with a high accuracy rate, that can ultimately change how weeds affect the current state of agriculture.

Keywords: Augmentation, CNN, dropout, plant seedling, ResNet

7.1 Introduction

The production of best featured quality seedlings is important if yields are to be improved and quality achieved. Plant seedling ...