Publication
Continuous object orientation estimation of cotton crops using deep convolutional neural networks
Conference Article
Conference
Catalan Conference on Artificial Intelligence, poster session (CCIA-P)
Edition
2025
Pages
270-274
Doc link
http://dx.doi.org/10.3233/FAIA250602
File
Abstract
Deep Convolutional Neural Networks (DCNNs) have shown an interesting performance in accurately estimating the continuous orientation of objects, successfully predicting angles ranging from 0 to 360 degrees based on RGB 2D image inputs to enhance manipulation. However, these types of solutions have not been explored for agriculture scenarios. In this study, we propose a real-world application using previously presented methods to estimate cotton crop orientation, which provides key information for a Next Best View (NBV) algorithm, enhancing decision-making with respect to the selection for the next optimal end-effector (EE) position for efficient and successful harvesting.
Categories
active vision, manipulators, object detection, planning (artificial intelligence), robot vision.
Scientific reference
A. Martínez and S. Foix. Continuous object orientation estimation of cotton crops using deep convolutional neural networks, 2025 Catalan Conference on Artificial Intelligence, poster session, 2025, Valls, Spain, pp. 270-274, IOS Press.
Follow us!