学习迁移
分类
人工智能
计算机科学
深度学习
机器学习
集合(抽象数据类型)
多样性(控制论)
人工神经网络
功能(生物学)
机制(生物学)
生物
哲学
认识论
进化生物学
程序设计语言
作者
Parveen Malik,Arunima K. Singh,Chiranjeet Gorai,I.S. Jha,Subhadip Paul
标识
DOI:10.1109/silcon59133.2023.10404367
摘要
Ecosystems cannot function without the presence of insects, yet due to their enormous diversity and minute variances, recognising and categorising them may be difficult. It is possible to reliably categorise insect species using deep learning algorithms. Deep learning-based insect categorization may be used for a variety of purposes, such as advancing scientific research, enhancing agricultural pest management, and assisting environmental conservation. The development of a deep learning model allows us to rapidly and precisely identify insects, even when they are in different life stages or seem identical to other species. By doing so, their actions and effects on the environment may be better understood. Considering the advancements in the fields of AI, especially in the area of deep learning, we have proposed to use transfer learning mechanism to classify the insects on IP102 dataset. Various standard models like VGG, Inception, Xception, MobileNet, DenseNet, ResNet and EfficientNet are used as a base model and then fine tuned to learn the representations. The best training set and validation set accuracy are found to be 71.34% and 64.95% respectively for the EfficientNetB7 model.
科研通智能强力驱动
Strongly Powered by AbleSci AI