Open AccessProceedings Article
Machine learning for adaptive image interpretation
Ilya Levner,Vadim Bulitko +1 more
- pp 870-876
TLDR
This paper proposes and implements several extensions of ADORE addressing its primary limitations that enable the first successful application of this emerging AI technology to a natural image interpretation domain and is shown to be robust with respect to noise in the training data, illumination, and camera angle variations.Abstract:
Automated image interpretation is an important task with numerous applications. Until recently, designing such systems required extensive subject matter and computer vision expertise resulting in poor cross-domain portability and expensive maintenance. Recently, a machine-learned system (ADORE) was successfully applied in an aerial image interpretation domain. Subsequently, it was re-trained for another man-made object recognition task. In this paper we propose and implement several extensions of ADORE addressing its primary limitations. These extensions enable the first successful application of this emerging AI technology to a natural image interpretation domain. The resulting system is shown to be robust with respect to noise in the training data, illumination, and camera angle variations as well as competitively adaptive with respect to novel images.read more
Citations
More filters
Journal ArticleDOI
Classification-Driven Watershed Segmentation
Ilya Levner,Hong Zhang +1 more
TL;DR: This paper presents a novel approach for creation of topographical function and object markers used within watershed segmentation that uses two classifiers, one trained to produce markers, the other training to produce object boundaries.
Proceedings ArticleDOI
Semi-supervised classification with hybrid generative/discriminative methods
TL;DR: It is conjecture that hybrid generative/discriminative methods allow semi-supervised learning in the presence of strongly overlapping classes and reduce the risk of modeling structure in the unlabeled data that is irrelevant for the specific classification task of interest.
Journal ArticleDOI
Habitat-Net: Segmentation of habitat images using deep learning
Jesse F. Abrams,Anand Vashishtha,Seth T. Wong,An Nguyen,Azlan Mohamed,Sebastian Wieser,Arjan Kuijper,Andreas Wilting,Anirban Mukhopadhyay +8 more
TL;DR: Habitat-Net is a novel deep learning application based on Convolutional Neural Networks to segment habitat images of tropical rainforests and reduces the degree of uncertainty that would be introduced by manual processing of images by different people (either over time or between study sites).
Journal ArticleDOI
3D Scene interpretation by combining probability theory and logic: The tower of knowledge
Mai Xu,Maria Petrou +1 more
TL;DR: The recursive version of ToK proves to be able to cope very well even without any training data, where it learns the characteristics of the various components by simply applying the pre-programmed logic rules that connect labels, actions and attributes.
References
More filters
Book
Reinforcement Learning: An Introduction
TL;DR: This book provides a clear and simple account of the key ideas and algorithms of reinforcement learning, which ranges from the history of the field's intellectual foundations to the most recent developments and applications.
Book
Introduction to Reinforcement Learning
TL;DR: In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning.
Journal ArticleDOI
Local Maximum Filtering for the Extraction of Tree Locations and Basal Area from High Spatial Resolution Imagery
TL;DR: In this article, the authors investigated the use of local maximum (LM) filtering to locate trees on high spatial resolution (1-m) imagery, in terms of commission error (falsely indicated trees) and omission error (missed trees).
Journal ArticleDOI
TIDA: an algorithm for the delineation of tree crowns in high spatial resolution remotely sensed imagery
TL;DR: TIDA was developed for application to imagery of native Eucalypt forests in Australia, and uses a 'top-down' spatial clustering approach involving key steps designed to reduce the effects of crown segmentation.