Deep Residual Networks for Plant Identification

Vinit Bodhwani, D. P. Acharjya, Umesh Bodhwani

Procedia Computer Science · 2019 · 71 citations · 19 references

DOIFull text

Open access

Concepts

Abstract

Advancing the knowledge and understanding of plants growing around us plays a very important and crucial role in medicinally, economically and in a sustainable agriculture. The identification of plant images in the field of computer vision has become an interdisciplinary focus. Taking benign conditions of quick advancement in computer vision and deep learning algorithms, convolutional neural networks (CNN) approach is taken into consideration to learn feature representation of 185 classes of leaves taken from Columbia University, the University of Maryland, and the Smithsonian Institution. For the large-scale classification of plants in the natural environment, a 50-layer deep residual learning framework consisting of 5 stages is designed. The proposed model achieves a recognition rate of 93.09 percent as the accuracy of testing on the LeafSnap dataset, illustrating that deep learning is a highly promising forestry technology.

References

19