Title:

APPLYING A DEEP LEARNING ALGORITHM TO THE PROBLEM OF CROP LEAF DISEASE

Author:

Rajasree R and Dr. Lilit Kumar Khatri

Abstract:

Agriculture requires plant disease detection to boost productivity. Recent image processing breakthroughs make visual plant disease analysis novel. Few papers, much alone in-depth studies, discuss this issue. Research explores the challenge of visual plant disease detection. Plant disease photos sometimes include several symptoms, irregularly distributed lesions, and different backgrounds, making them hard to recognize. We created a plant disease collection with 220,592 images and 271 categories for identification research. Reweighting visual regions and loss uncovers plant illnesses in this dataset. We use cluster distribution to calculate the weights of all split patches from each image to determine each patch's discriminative level. To improve discriminative sickness component learning in weakly-supervised training, we weight each loss per pair of patches. We encode the weighted patch feature sequence into a complete feature representation using the LSTM network after extracting patch features from the loss reweighted network. Significant testing on this and other public datasets shows the recommended technique is superior. Our results should help image processing detect plant diseases

Keyword:

Deep learning, Active Contour Method, Convolutional Neural Network.