Saliency-Guided Unsupervised Feature Learning for Scene Classification

By Fan Zhang, Bo Du, Liangpei Zhang. 2015

528 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology, Ocean Engineering, Representation & Self-Supervised Learning

How researchers see this paper

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

In Computer Vision and Pattern Recognition

1. Junwei Han
2. Gong Cheng
3. Liangpei Zhang
4. Dacheng Tao
5. Bo Du
6. Xiaoqiang Lu
7. Huchuan Lu
8. G. Wang

In Media Technology

1. Liangpei Zhang
2. Gong Cheng
3. Junwei Han
4. Xiaoqiang Lu
5. Bo Du
6. Gui-Song Xia
7. Qian Du
8. Yanfei Zhong

In Ocean Engineering

1. Gong Cheng
2. Junwei Han
3. Liangpei Zhang
4. Xiaoqiang Lu
5. Vladimir Risojevic
6. Xiwen Yao
7. Gui-Song Xia
8. Qian Du

In Representation & Self-Supervised Learning

1. Liangpei Zhang
2. Gui-Song Xia
3. Yanfei Zhong
4. Fan Hu
5. Jingwen Hu
6. Bo Du
7. Xiaoqiang Lu
8. Gong Cheng

Citations in the corpus

• Remote Sensing Image Scene Classification: Benchmark and State of the Art (2017, methodology): “unsupervised feature learning method that has been successfully applied to land use scene classification [63, 129, 133, 134].”
• Hierarchical Coding Vectors for Scene Level Land-Use Classification (2016, unknown): “…Structural texture similarity [33] 86.0 Wavelet BOVW [29] 87.4 ˘ 1.3 Unsupervised feature learning [34] 81.1 ˘ 1.2 Saliency-guided feature learning [35] 82.7 ˘ 1.2 Concentric circle-structured BOVW [2] 86.6 ˘ 0.8 Multifeature concatenation [36] 89.5 ˘ 0.8 Pyramid-of-spatial-relations [3] 89.1 CLBP…”
• Deep Feature Fusion for VHR Remote Sensing Scene Classification (2017, methodology): “We then compare the scene classification performance of the proposed method with 1) different approaches based on handcrafted features including SIFT and SC approach cited in [6], saliency guided sparse autoencoder noted in [8], recent work”
• Accurate Object Localization in Remote Sensing Images Based on Convolutional Neural Networks (2017, background): “feature transform (SIFT) [1], histogram of oriented gradients (HOG) [2], and Saliency [3], [4].”
• Building Instance Classification Using Street View Images (2018, unknown): “, 2013) and features learned by sparse coding methods (Wang et al., 2014; Yang et al., 2014; Sun et al., 2015; Rigas et al., 2013; Zhang et al., 2015; Tuia et al., 2015, 2016; Yao et al., 2016; Cheng et al., 2015).”
• A Stagewise Refinement Model for Detecting Salient Objects in Images (2017, background): “Salient object detection models have gained broad interest recently due to their applications in several fields such as tracking [19], image understanding [54, 45], person re-identification [4], image captioning [51, 13, 12], visual question answering [32], and object proposal generation [2].”
• Exploring Models and Data for Remote Sensing Image Caption Generation (2017, methodology): “2) Sydney-Captions Data Set: This data set is also provided by [26], which is based on the Sydney data set [42].”
• Scene Classification via a Gradient Boosting Random Convolutional Network Framework (2016, result): “Finally, we also compared the scene classification performance of the proposed approach with the spatial pyramid matching kernel (SPMK) [3], the SIFT + sparse coding (SSC) approach described in [1], and the saliency-guided sparse autoencoder (SSAE) method in [32].”
• A Comparative Study of Sampling Analysis in the Scene Classification of Optical High-Spatial Resolution Remote Sensing Imagery (2015, methodology): “Some work in the scene classification of optical HSR-RS images has attempted to use saliency-based sampling [13], but it is still not clear how it can help the scene classification in other feature learning methods compared to other strategies.”
• Multiscale Dynamic Graph Convolutional Network for Hyperspectral Image Classification (2019, background): “To address this defect, deep learning [15], [16], [17], [18] has been extensively employed for hyperspectral image classifi-cation and has attracted increasing attention for its strong representation ability.”
• High-resolution remote-sensing imagery retrieval using sparse features by auto-encoder (2015, methodology): “In a recent work by Zhang, Du, and Zhang (2015), the sparse auto-encoder network, an unsupervised feature learning method, was trained with both salient and non-salient patches to learn feature representations for aerial scene classification and achieved better performance than the approach…”
• AID: A Benchmark Data Set for Performance Evaluation of Aerial Scene Classification (2016, background): “have been developed to automatically learn adaptive feature representations from images [20], [32], [37].”
• Transferring Deep Convolutional Neural Networks for the Scene Classification of High-Resolution Remote Sensing Imagery (2015, methodology): “[39] used a classic neural network, called sparse autoencoder, which is trained on a group of selective image patches sampled by their saliency degree, to extract local features.”
• Unsupervised Feature Learning Via Spectral Clustering of Multidimensional Patches for Remotely Sensed Scene Classification (2015, methodology): “[54] present a saliency-guided UFL method in which the standard UFL pipeline is explored with the sparse autoencoder (an unsupervised symmetrical neural network) used to”
• Deep Learning Earth Observation Classification Using ImageNet Pretrained Networks (2016, background): “Authors in [6] and [7] have investigated such an approach over RS data; however, their implementations remain shallow due to the use of a single feature extraction level.”
• Deep semantic understanding of high resolution remote sensing image (2016, methodology): “…a HSR remote sensing image, it is necessary to consider maximizing the probability where b 0 is the image feature vector, g ( · ) is computational process of RNN or LSTMs. Note that w 1 is a special START vector and w N is an END vector, which represent the start and end of the sentence,…”