Article
Computer Science, Information Systems
Cheng Peng, Yikun Liu, Xinpan Yuan, Qing Chen
Summary: In order to improve the accuracy of CNN in image classification, an enhanced Inception-ResNet-v2 model based on CNN is designed by comparing and analyzing the structure of classification models. The use of multi-scale depthwise separable convolution reduces the amount of model parameters and extracts features under different receptive fields. The establishment of a channel filtering module based on global information comparison enables effective feature extraction, and the model achieves better accuracy than most other models in each dataset, with an accuracy rate of 94.8%.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Qing Zhang, Yong Yan, Yong Lin, Yan Li
Summary: This study proposes an image retrieval method based on deep learning, which achieves secure retrieval of ciphertext images through a deep artificial neural network model and an image encryption algorithm. Experimental results show significant improvements in retrieval efficiency and performance indicators, highlighting the importance of research on image information security retrieval.
Article
Computer Science, Artificial Intelligence
Wei Chen, Yu Liu, Weiping Wang, Erwin M. Bakker, Theodoros Georgiou, Paul Fieguth, Li Liu, Michael S. Lew
Summary: In recent years, there has been a massive increase in the generation and sharing of visual content in various fields, such as social media platforms, medical imaging, and robotics. This has created new challenges, particularly in searching databases for similar content using Content Based Image Retrieval (CBIR). Artificial intelligence has made progress in CBIR through deep learning algorithms and techniques, improving efficiency and accuracy in real-time retrieval. This survey reviews recent works in instance retrieval based on deep learning, covering deep feature extraction, feature embedding and aggregation methods, and network fine-tuning strategies.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Plant Sciences
Jin Liang, Wenping Jiang
Summary: This paper proposes a ResNet50-DPA model for tomato leaf disease identification. By introducing an improved ResNet50 and a dual-path attention mechanism, it can capture key features more accurately and improve the accuracy of disease identification. In addition, incorporating the DPA module into the residual module can also reduce economic losses.
FRONTIERS IN PLANT SCIENCE
(2023)
Article
Computer Science, Interdisciplinary Applications
Lin Ge, Xingyue Wei, Yayu Hao, Jianwen Luo, Yan Xu
Summary: This study proposed an unsupervised structural feature guided convolutional neural network method for the registration of multiple stained images. Through the combination of low-resolution and high-resolution structural features, as well as a multi-scale strategy, it effectively overcame challenges such as repetitive texture and section missing.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2022)
Article
Chemistry, Analytical
Muhammad Mostafa Monowar, Md Abdul Hamid, Abu Quwsar Ohi, Madini O. Alassafi, M. F. Mridha
Summary: Image retrieval techniques are gaining popularity due to the availability of multimedia data. This paper introduces AutoRet, a self-supervised image retrieval system based on deep convolutional neural networks (DCNN). The system is trained on pairwise constraints and can work with partially labeled datasets. Benchmarking results show that the proposed method performs well in a self-supervised manner and can handle mixed availability of labeled data.
Article
Computer Science, Information Systems
Xingxu Yao, Dongyu She, Haiwei Zhang, Jufeng Yang, Ming-Ming Cheng, Liang Wang
Summary: The paper introduces a method for processing affective images through adaptive deep metric learning, which enhances the recognition of emotional images by designing adaptive sentiment similarity loss and sentiment vector, while also proposing a unified multi-task deep framework.
IEEE TRANSACTIONS ON MULTIMEDIA
(2021)
Article
Engineering, Electrical & Electronic
Yameng Wang, Shunping Ji, Yongjun Zhang
Summary: This paper introduces a learnable joint spatial and spectral transformation (JSST) model for remote sensing image retrieval, which adaptsively learns geometric and spectral transformation parameters through a parameter generation network to achieve geometric and spectral correction of images, thus enhancing the generalization and adaptation ability of cross-dataset remote sensing image retrieval.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2021)
Article
Engineering, Biomedical
Rohollah Hedayati, Mohammad Khedmati, Mehran Taghipour-Gorjikolaie
Summary: Alzheimer's disease is a major cause of death among the elderly, but early diagnosis is difficult. Machine learning methods have been used to improve accuracy in diagnosing Alzheimer's disease, with results showing accuracy rates of 95% for AD/NC, 90% for AD/MCI, and 92.5% for MCI/NC. This indicates the method is reliable for early diagnosis of Alzheimer's disease.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2021)
Article
Computer Science, Artificial Intelligence
Toshi Sinha, Brijesh Verma
Summary: This study proposes a novel architecture of convolutional neural network with a non-iterative radial basis function-based classification layer, making it more efficient for image classification tasks. Experimental results show that the proposed architecture outperforms the traditional CNN methods in terms of accuracy, parameter tuning, and time efficiency.
APPLIED SOFT COMPUTING
(2022)
Article
Computer Science, Artificial Intelligence
Pawel Staszewski, Maciej Jaworski, Jinde Cao, Leszek Rutkowski
Summary: In this brief, a novel algorithm is proposed for constructing effective descriptors for content-based image retrieval using deep neural networks. The descriptors made up of values from both fully connected and convolutional layers perfectly represent the entire image content. Experimental verification showed the effectiveness of these descriptors in semantic matching and secondary image characteristics.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Geochemistry & Geophysics
Xunyang Su, Jinjiang Li, Zhen Hua
Summary: In this paper, a transformer-based regression network (DR-NET) architecture is proposed for pansharpening in remote sensing. The DR-NET consists of feature extraction, feature fusion, and image reconstruction stages to obtain images with uniform spectral distribution and rich spatial details. Experimental results demonstrate the superior performance of DR-NET.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Komal Nain Sukhia, Syed Sohaib Ali, M. Mohsin Riaz, Abdul Ghafoor, Benish Amin
Summary: This letter introduces a content-based image retrieval technique using a novel dense angle descriptor and dictionary learning. It addresses the issue of rotation invariance in image retrieval by presenting a dense angle-based HOG descriptor. Experimental results on different datasets demonstrate that the proposed technique achieves high retrieval performance.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Donggen Li, Dawei Dai, Jiancu Chen, Shuyin Xia, Guoyin Wang
Summary: Deep hashing combines feature extraction or representation with hash coding jointly, which can significantly improve the speed of large-scale image retrieval. However, the retrieval performance of binary hash coding has declined to a certain extent due to the reduction of dimension and information loss compared with traditional retrieval methods.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Zhedong Zheng, Liang Zheng, Yi Yang, Fei Wu
Summary: This paper aims to generate adversarial queries for image retrieval by directly attacking query image features. The proposed opposite-direction feature attack (ODFA) method induces the original image feature in the opposite direction to create adversarial queries. Experimental results on five retrieval datasets demonstrate that ODFA outperforms classifier attack methods in terms of attack success rate, causing true matches to be less visible in the top ranks. Moreover, the method is extendable to multi-scale query inputs and applicable in black-box settings.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2023)