Multiview Deep Learning for Land-Use Classification

By Francois P. S. Luus, Brian Paxton Salmon, Frans van den Bergh, Bodhaswar Tikanath Jugpershad Maharaj. 2015

245 citations recorded by the source corpus.

A multiscale input strategy for multiview deep learning is proposed for supervised multispectral land-use classification, and it is validated on a well-known data set. The hypothesis that simultaneous multiscale views can improve composition-based inference of classes containing size-varying objects compared to single-scale multiview is investigated. The end-to-end learning system learns a hierarchical feature representation with the aid of convolutional layers to shift the burden of feature determination from hand-engineering to a deep convolutional neural network (DCNN). This allows the classifier to obtain problem-specific features that are optimal for minimizing the multinomial logistic regression objective, as opposed to user-defined features which trade optimality for generality. A heuristic approach to the optimization of the DCNN hyperparameters is used, based on empirical performance evidence. It is shown that a single DCNN can be trained simultaneously with multiscale views to improve prediction accuracy over multiple single-scale views. Competitive performance is achieved for the UC Merced data set, where the 93.48% accuracy of multiview deep learning outperforms the 85.37% accuracy of SIFT-based methods and the 90.26% accuracy of unsupervised feature learning.

Subfields: Atmospheric Science, Computer Vision and Pattern Recognition, Global and Planetary Change, Media Technology, Multimodal Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Atmospheric Science

In Computer Vision and Pattern Recognition

In Global and Planetary Change

In Media Technology

In Multimodal Learning

Citations in the corpus

EuroSAT: A Novel Dataset and Deep Learning Benchmark for Land Use and Land Cover Classification (2017, background): “As shown in previous work [6], [15], [17], [19], deep CNNs have been shown to outperform non-deep learning approaches in land use and land cover image classification.”
In-domain representation learning for remote sensing (2019, background): “Because it was not possible to train large state-of-the-art network on a small dataset from scratch, remote sensing researchers started to build smaller networks, as for instance in Luus et al. (2015).”
Polarimetric SAR Image Classification Using Deep Convolutional Neural Networks (2016, methodology): “It has been applied to remote sensing image processing [10], [11] and classification [12], [13], while convolutional neural networks as one of the typical deep learning models have achieved impressive performances in various fields [11], [13].”
Deep Learning for Remote Sensing Data: A Technical Tutorial on the State of the Art (2016, methodology): “suggested a multiscale input strategy for multiview DL with the aid of convolutional layers to shift the burden of feature determination from hand-engineering to a deep CNN [124].”
Deep learning for remote sensing image classification: A survey (2018, methodology): “There have been different approaches proposed in the literature, where smaller networks are used to better fit the RS images (Luus et al., 2015; Volpi & Tuia, 2016; Zhang, Du, & Zhang, 2016; Zou et al., 2015), which may also help improve the performance regarding the overfitting and local minimum…”
Deep learning in remote sensing: a review (2017, background): “Consequently, some construct a new smaller network and train it from scratch using satellite images to better fit the satellite data [80], [84], [85], [92].”
Deep learning for change detection in remote sensing: a review (2022, background): “…(i.e. flips, mirror, rotations, and cropping) for training samples diversifies its holistic spatial layout and orientation (Yu et al. 2017), which can effectively avoid the over-fitting problem in deep learning (Cireşan et al. 2010; Luus et al. 2015; Simard, Steinkraus, and Platt 2003).”
AID: A Benchmark Data Set for Performance Evaluation of Aerial Scene Classification (2016, methodology): “This type of methods also achieves the state-of-the-art performance on the aerial scene analysis [4], [33]–[35], [38], [41], [42], [88], [89].”
Convolutional Neural Networks for Large-Scale Remote-Sensing Image Classification (2017, unknown): “Let us remark that convolutional neural networks have also been used for other remote sensing applications, such as road tracking [25], object detection [26] and land use classiﬁcation [27], [28].”
Deep learning in agriculture: A survey (2018, background): “CNN also seem to be superior than unsupervised feature learning with 3–11% higher CA (Luus et al., 2015), 2–44% improved CA in relation to local shape and color features (Dyrmann et al.”
Training Deep Convolutional Neural Networks for Land–Cover Classification of High-Resolution Imagery (2017, background): “[14] created a deep network from scratch and achieved a 93.”
Classification of Remote Sensing Images Using EfficientNet-B3 CNN Model With Attention (2021, background): “A large amount of works that deals with this problem is published in the RS literature [9]–[29].”