2021 · 27 citations · 7 references
ComputingAgronomyComputer ScienceComputer VisionMachine LearningPlant NutritionCrop NutritionDeep LearningNeural NetworksNutrient Deficiency DiagnosisNutrient Deficiency SymptomsReinforcement LearningConvolutional Neural NetworksDeep Learning-based ClassificationDeep Reinforcement LearningNitrogen Deficiency SymptomsNutrient Deficiency ClassificationNutrient Deficiency DetectionNutrient Deficiency PhenotypingNutrient Deficiency Symptom Recognition
Rice is an important crop for agricultural production. Because of the vulnerability of rice, it is likely to be affected by many external conditions. Particularly, rice may lack of various kinds of nutrition elements such as potassium, nitrogen, and phosphorus. For people who are not experienced in botany, it is hard for them to identify nutrition deficiency and add corresponding supplement. Symptoms of nutrient deficiencies in rice plants often are usually represented by phenotype of the leaves. Therefore, the phenotype of leaves can be examined to identify nutrition deficiency. Our goal is to utilize deep learning, specifically, convolutional neural network (CNN), and combine reinforcement learning to help people to address this problem. Experiments were conducted with a dataset containing 1,500 images of rice leaves subject to three different types of nutrition deficiency—nitrogen (N), phosphorus (P), and potassium (K) deficiencies. This research preprocesses images first, constructs different CNN architectures, trains the model, and compares the results. With the reinforcement learning augmentation, experiments indicate that Densenet-121 is the best deep CNN model among these structures we had trained to identify what nutrition does a plant’s leave lack of with a test accuracy of 97%. This study demonstrates Densenet-121 model is a good tool to diagnose nutrient deficiency in crops and attempts to incorporate reinforcement learning to augment inputs.
7
Deep Residual Learning for Image Recognition
Kaiming He, Xiangyu Zhang, Shaoqing Ren et al. · 2016 · 222.1K citations · Full text
Going deeper with convolutions
Christian Szegedy, Wei Liu, Yangqing Jia et al. · 2015 · 46.8K citations
Densely Connected Convolutional Networks
Gao Huang, Zhuang Liu, Laurens van der Maaten et al. · 2017 · 44.7K citations
Streetscore -- Predicting the Perceived Safety of One Million Streetscapes
Nikhil Naik, Jade Philipoom, Ramesh Raskar et al. · 2014 · 452 citations · Full text