Meta-Hashing for Remote Sensing Image Retrieval

Tang, Xu; Yang, Yahan; Ma, Jingjing; Cheung, Yiu-ming; Liu, Chao; Liu, Fang; Zhang, Xiangrong; Jiao, Licheng

doi:10.1109/tgrs.2021.3136159

Cited by 36 publications

(20 citation statements)

References 55 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…CA-SVDD could achieve the highest performance for classes: river (5), sparse residential (8), and baseball diamonds (11). In contrast, GCA-ResNet and CBP-ResNet perform quite similarly for most classes and both methods have shown the best performance for classes: beach (1), mobile home park (14), intersection (15), and tennis court (21). TCAL has the highest performance for some classes such as chaparral (12), and mobile home (14).…”

Section: Comparison With Prior Workmentioning

confidence: 96%

“…However, this method needs a large amount of supervised information for training. Later, to overcome this limitation, they proposed metahashing [21] by developing a self-adaptive convolution block that could achieve high retrieval performance with a few labeled training samples. 3) Cross-modal RS image retrieval [22], [23], [24]: This category takes advantage of other kinds of data sources.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Query by Example in Remote Sensing Image Archive Using Enhanced Deep Support Vector Data Description

Ghozatlou

Conde

Datcu

2023

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

This article studies remote sensing image retrieval using kernel-based support vector data description (SVDD). We exploit deep SVDD, which is a well-known method for one-class classification to recover the most relevant samples from the archive. To this end, a deep neural network (DNN) is jointly trained to map the data into a hypersphere of minimum volume in the latent space. It is expected that similar samples to the query are compressed inside of the hypersphere. The closest embedding to the center of the hypersphere is related to the most relevant sample to query. We enhance deep SVDD by injecting the statistical information of data to the DNN by means of additional terms in the cost function. The first enhancement method takes advantage of covariance regularization of batches of the training set to penalize unnecessary redundancy and minimize the correlation between the different dimensions of the embedding. The second method involves unlocking the hypersphere's predefined center while preventing network divergence during training. Therefore, two parameters are designed to control the importance of the drifting of the center and the importance of a fixed predefined center (convergence), respectively. This has been implemented by considering the average of batches of embedding in each iteration as the updated center. This pushes irrelevant samples away from query samples, making data clustering easier for the DNN. The performance of the proposed methods is evaluated on benchmark datasets.

show abstract

Section: Comparison With Prior Workmentioning

confidence: 96%

mentioning

confidence: 99%

Query by Example in Remote Sensing Image Archive Using Enhanced Deep Support Vector Data Description

Ghozatlou

Conde

Datcu

2023

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Feature extraction/learning is of vital importance for US-CBRSIR. Numerous feature descriptors, ranging from hand-crafted features [13] to deep-learning-based features [14,15], are exploited and applied to map the HRRS images into discriminative features. Then, simple or specific distance metrics [16] are designed to complete US-CBRSIR according to the resemblance between features.…”

Section: Introductionmentioning

confidence: 99%

Dual Modality Collaborative Learning for Cross-Source Remote Sensing Retrieval

Shi

Tang

et al. 2022

Remote Sensing

Self Cite

View full text Add to dashboard Cite

Content-based remote sensing (RS) image retrieval (CBRSIR) is a critical way to organize high-resolution RS (HRRS) images in the current big data era. The increasing volume of HRRS images from different satellites and sensors leads to more attention to the cross-source CSRSIR (CS-CBRSIR) problem. Due to the data drift, one crucial problem in CS-CBRSIR is the modality discrepancy. Most existing methods focus on finding a common feature space for various HRRS images to address this issue. In this space, their similarity relations can be measured directly to obtain the cross-source retrieval results straight. This way is feasible and reasonable, however, the specific information corresponding to HRRS images from different sources is always ignored, limiting retrieval performance. To overcome this limitation, we develop a new model for CS-CBRSIR in this paper named dual modality collaborative learning (DMCL). To fully explore the specific information from diverse HRRS images, DMCL first introduces ResNet50 as the feature extractor. Then, a common space mutual learning module is developed to map the specific features into a common space. Here, the modality discrepancy is reduced from the aspects of features and their distributions. Finally, to supplement the specific knowledge to the common features, we develop modality transformation and the dual-modality feature learning modules. Their function is to transmit the specific knowledge from different sources mutually and fuse the specific and common features adaptively. The comprehensive experiments are conducted on a public dataset. Compared with many existing methods, the behavior of our DMCL is stronger. These encouraging results for a public dataset indicate that the proposed DMCL is useful in CS-CBRSIR tasks.

show abstract

“…Then, image retrieval can be achieved by calculating the Hamming distances with simple bitwise XOR operations [5]. Several hashing methods are presented in RS [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16]. The traditional hashing methods extract hand-crafted image features and map them into low-dimensional binary codes by using hashing functions [6], [7], [8].…”

mentioning

confidence: 99%

“…In [14], a generative adversarial network is exploited for hash code learning, while similar losses to DCHHN for the generator and a sigmoid function for the discriminator are used to determine whether the generated codes are true codes that comply with the bit-balancing rule. In [15], a meta-hashing algorithm is introduced to increase the generalization capability of DNNs utilized for hash code generation under a small number of training samples. To this end, this algorithm employs few-shot meta-learning for hash code generation by dividing a learning objective into multiple subtasks and using all training samples multiple times.…”

mentioning

confidence: 99%

Towards Simultaneous Image Compression and Indexing for Scalable Content-Based Retrieval in Remote Sensing

Sümbül

Xiang

Demir

2022

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

Due to the rapidly growing remote-sensing (RS) 1 image archives, images are usually stored in a compressed format 2 for reducing their storage sizes. Thus, most of the existing 3 content-based RS image retrieval systems require fully decoding 4 images (i.e., decompression) that is computationally demanding 5 for large-scale archives. To address this issue, we introduce a 6 novel approach devoted to simultaneous RS image compres-7 sion and indexing for scalable content-based image retrieval 8 (denoted as SCI-CBIR). The proposed SCI-CBIR prevents the 9 requirement of decoding RS images before image search and 10 retrieval. To this end, it includes two main steps: 1) deep-11 learning-based compression and 2) deep-hashing-based indexing. 12 The first step effectively compresses RS images by employing 13 a pair of deep encoder and decoder neural networks and an 14 entropy model. The second step produces hash codes with a high 15 discrimination capability for RS images by employing pairwise, 16 bit-balancing, and classification loss functions. For the training 17 of the SCI-CBIR approach, we also introduce a novel multistage 18 learning procedure with automatic loss weighting techniques 19 to characterize RS image representations that are appropriate 20 for both RS image indexing and compression. The proposed 21 learning procedure enables automatically weighting different 22 loss functions considered for the proposed approach instead of 23 computationally demanding grid search. Experimental results 24 show the effectiveness of the proposed approach when compared 25 to widely used approaches in RS. The code of the proposed 26 approach is available at https://git.tu-berlin.de/rsim/SCI-CBIR. 27 Index Terms-Deep-learning-based compression, hashing-28 based indexing, image retrieval, remote sensing (RS). 29 I. INTRODUCTION 30 R ECENT advances in satellite technologies lead to a 31 significantly increased volume of remote-sensing (RS) 32 image archives. Thus, in recent years, increasing attention 33 has been devoted to the development of accurate and scal-34 able content-based image retrieval (CBIR) methods for such 35 archives [1], [2], [3]. For large-scale CBIR, fast and accurate 36 indexing methods that allow approximate nearest neigh-37 bor search are fundamental. In this perspective, hashing-38 based indexing has recently attracted attention to solving 39

show abstract

Meta-Hashing for Remote Sensing Image Retrieval

Cited by 36 publications

References 55 publications

Query by Example in Remote Sensing Image Archive Using Enhanced Deep Support Vector Data Description

Query by Example in Remote Sensing Image Archive Using Enhanced Deep Support Vector Data Description

Dual Modality Collaborative Learning for Cross-Source Remote Sensing Retrieval

Towards Simultaneous Image Compression and Indexing for Scalable Content-Based Retrieval in Remote Sensing

Contact Info

Product

Resources

About