Abstract: The agricultural industry, pivotal for global food security and sustainability, grapples with a persistent challenge posed by insect pests wreaking havoc on crops. Identifying these pests and discerning their maturity stages are crucial for effective pest management and safeguarding crop yields. Traditional manual identification methods prove imprecise, time-consuming, and often inefficient, even for seasoned agronomists, due to visual similarities among insect species, especially at identical maturity stages. Notably, deep learning emerges as a prominent approach, albeit demanding extensive labeled datasets for effective training, a resource that remains scarce, demanding, and insufficient in addressing the wide-ranging variability within insect classes. Additionally, integrating pesticide recommendation systems could enhance pest management strategies, aiding in the selection of appropriate treatments for specific pests and crop types. This research proposes a solution to this problem using a few-shot learning approach as a solution to this predicament, delving into insect classification for pest management. A few-shot prototypical network is proposed based on a comparison with other state-of-art models and divergence analysis. Experiments were conducted separating the adult classes and the early stages into different groups, and at last recommending suitable pesticides that will help in th yeilding of good crops.
Keywords: Few-shot learning; Insect pest classification; Insect maturity stages; Convulution Neural Network; Prototypical Networks.
Cite:
Birajdar Siddanna Gurabala, Saloni, Shreya Shetty, Varshitha G V, Ms. Sunitha N V, "INSECT PEST IMAGE RECOGNITION AND MATURITY STAGES CLASSIFICATION USING FEW-SHOT MACHINE LEARNING APPROACH", IJARCCE International Journal of Advanced Research in Computer and Communication Engineering, vol. 13, no. 3, 2024, Crossref https://doi.org/10.17148/IJARCCE.2024.13373.
|
DOI:
10.17148/IJARCCE.2024.13373
