Self‐supervised learning improves classification of agriculturally important insect pests in plants
نویسندگان
چکیده
Insect pests cause significant damage to food production, so early detection and efficient mitigation strategies are crucial. There is a continual shift toward machine learning (ML)-based approaches for automating agricultural pest detection. Although supervised has achieved remarkable progress in this regard, it impeded by the need expert involvement labeling data used model training. This makes real-world applications tedious oftentimes infeasible. Recently, self-supervised (SSL) have provided viable alternative training ML models with minimal annotations. Here, we present an SSL approach classify 22 insect pests. The framework was assessed on raw segmented field-captured images using three different methods, Nearest Neighbor Contrastive Learning of Visual Representations (NNCLR), Bootstrap Your Own Latent, Barlow Twins. pre-training done ResNet-18 ResNet-50 all methods original RGB foreground images. performance evaluated linear probing representations end-to-end fine-tuning approaches. SSL-pre-trained convolutional neural network were able perform annotation-efficient classification. NNCLR best performing method both full fine-tuning. With just 5% annotated images, transfer ImageNet initialization obtained 74% accuracy, whereas improved classification accuracy 79% Models created consistently performed better, especially under very low annotation, robust object class imbalances. These help overcome annotation bottlenecks resource efficient.
منابع مشابه
Survey Studies on Insect-pests Associated with Important Medicinal Plants in Himachal Pradesh
Indian Himalayan region is one of the richest sources of biodiversity in the world, having a wide range of medicinal and aromatic plants. The present studies were undertaken to record the insect-pest complex of some important medicinal plants. In all fifteen insect species belonging to four orders were observed to be associated with ten medicinal plants in different parts of Himachal Pradesh. A...
متن کاملOrigin and Spread of Important Insect Pests of Animals
As FAR BACK as the days of Pliny in the first century, and probably long before, horses, cows, sheep, goats, dogs, and chickens were known to be infested with fleas, bots, and lice. Until the white man arrived in America not one of these animals had been present on this continent for a long time, with the possible exception of the dog. It is natural to assume therefore that all of the specific ...
متن کاملComposition, distribution and economic importance of insect pests of prioritized aromatic plants in some growing of Ethiopia
Ethiopia is remarkably rich in its biological, ecological and landscape diversity and is home to outstanding natural bio-resources such as a number of herbs, aromatic and medicinal plants. The present study was carried out to study the composition, distribution and economic importance of insect pests of prioritized aromatic (Mentha sp., lemon verbena, fennel, rosemary, rosadamascene, oregano, p...
متن کاملComposition, distribution and economic importance of insect pests of prioritized aromatic plants in some growing of Ethiopia
Ethiopia is remarkably rich in its biological, ecological and landscape diversity and is home to outstanding natural bio-resources such as a number of herbs, aromatic and medicinal plants. The present study was carried out to study the composition, distribution and economic importance of insect pests of prioritized aromatic (Mentha sp., lemon verbena, fennel, rosemary, rosadamascene, oregano, p...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Plant phenome journal
سال: 2023
ISSN: ['2578-2703']
DOI: https://doi.org/10.1002/ppj2.20079