A survey on image classification of lightweight convolutional neural network
- Authors: Liu, Ying , Xiao, Peng , Fang, Jie , Zhang, Dengsheng
- Date: 2023
- Type: Text , Conference paper
- Relation: 19th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery, ICNC-FSKD 2023, Harbin, China, 29-31 July 2023, 2023 19th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery (ICNC-FSKD)
- Full Text: false
- Reviewed:
- Description: In recent years, deep neural networks have achieved tremendous success in image classification in both academic and industrial settings. However, the high hardware requirements imposed by their intensive and complex computations pose a challenge for deployment on low-storage devices. To address this challenge, lightweight networks provide a viable solution. This paper provides a detailed review of recent lightweight image classification algorithms, which can be categorized into low-redundancy network model design and neural network compression algorithms. The former reduces network computations by replacing traditional convolution with efficient lightweight convolution, while the latter reduces redundancy in the network by employing methods such as network pruning, knowledge distillation, and parameter quantization. We summarize the experimental results of some classical models and algorithms on ImageNet2012 and CIFAR-10 datasets, and analyze the characteristics, advantages and disadvantages of these models respectively. Finally, future research directions for lightweight algorithms in the field of image classification are identified. © 2023 IEEE.
Fine-grained image classification based on knowledge distillation
- Authors: Liu, Ying , Feng, Hao , Zhang, Weidong , Fang, Jie , Xiao, Peng , Zhang, Dengsheng
- Date: 2023
- Type: Text , Conference paper
- Relation: 19th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery, ICNC-FSKD 2023, Harbin, China, 29-31 July 2023, 2023 19th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery (ICNC-FSKD)
- Full Text: false
- Reviewed:
- Description: Despite the outstanding performance of deep learning-based fine-grained image classification methods, the commonly used models still suffer from high cost of computation and memory Therefore, this paper proposes a mobile-based CNN network that focuses on discriminative features of fine-grained images by embedding a hybrid-domain attention module to achieve higher accuracy in recognition. Specifically, under the premise of reducing network parameters, this paper presents a classification method that combines transfer learning and knowledge distillation to enhance the model's generalization performance and resistance to overfitting. Different knowledge transfer strategies are validated through the experiments in the knowledge distillation process. Mobile models such as SqueezeNet, MobileNetV2, and CBAM MobileNetV2 all demonstrate enhanced performance the knowledge distillation optimization. The proposed method in this paper can be used to develop a lightweight mobile-based CNN model with comparable performance to complex models making it more advantageous in real-life scenarios with limited storage resources and low hardware computation levels. Additionally, the model compression process utilizes only the intermediate features of the original dataset, meeting the confidentiality requirements of the original data in the field of public security. © 2023 IEEE.
A Rotation invariant HOG descriptor for tire pattern image classification
- Authors: Liu, Ying , Ge, Yuxiang , Wang, Fuping , Liu, Qiqi , Lei, Yanbo , Zhang, Dengsheng , Lu, Guojun
- Date: 2019
- Type: Text , Conference proceedings
- Relation: ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP); Brighton, UK, 12-17 May 2019. p. 2412-2416
- Full Text: false
- Reviewed:
- Description: Texture feature is important in describing tire pattern image which provides useful clue in solving crime cases and traffic accidents. In this paper, we propose a novel texture feature extraction method based on HOG (Histogram of Oriented Gradient) and dominant gradient (DG) in tire pattern images, named HOG-DG. The proposed HOG-DG is not only robust to illumination and scale changes but also is rotation-invariant. In the proposed HOG-DG, HOG features are first computed from circular local cells, and HOG features from an image are concatenated and normalized using the DG to construct the HOG-DG feature. HOG-DG is used to train a support-vector-machine (SVM) classifier for tire pattern classification. Experimental results demonstrate its outstanding performance for tire pattern description.
Improved kernel descriptors for effective and efficient image classification
- Authors: Karmakar, Priyabrata , Teng, Shyh , Zhang, Dengsheng , Liu, Ying , Lu, Guojun
- Date: 2017
- Type: Text , Conference proceedings
- Relation: 2017 International Conference on Digital Image Computing: Techniques and Applications (DICTA); Sydney, Australia; 29th November-1st December 2017 p. 195-202
- Full Text: false
- Reviewed:
- Description: Kernel descriptors have been proven to outperform existing histogram based local descriptors as such descriptors are extracted from the match kernels which measure similarities between image patches using different pixel attributes (gradient, colour or LBP pattern). The extraction of kernel descriptors does not require coarse quantization of pixel attributes. Instead, each pixel equally participates in matching between two image patches. In this paper, by leveraging the kernel properties, we propose a unique approach which simultaneously increases the effectiveness and efficiency of the existing kernel descriptors. Specifically, this is done by improving the similarity measure between two different patches in terms of any pixel attribute. The proposed kernel descriptors are more discriminant, take less time to be extracted and have much lower dimensions. Our experiments on Scene Categories and Caltech 101 databases show that our proposed approach outperforms the existing kernel descriptors.
Improved Tamura features for image classification using kernel based descriptors
- Authors: Karmakar, Priyabrata , Teng, Shyh , Zhang, Dengsheng , Liu, Ying , Lu, Guojun
- Date: 2017
- Type: Text , Conference proceedings
- Relation: 2017 International Conference on Digital Image Computing: Techniques and Applications (DICTA); Sydney, Australia; 29th November-1st December 2017 p. 461-467
- Full Text: false
- Reviewed:
- Description: Tamura features are based on human visual perception and have huge potential in image representation. Conventional Tamura features only work on homogeneous texture images and perform poor on generic images. Therefore, many researchers attempt to improve Tamura features and most of the improvements are based on histogram based representation. Kernel descriptors have been shown to outperform existing histogram based local features as such descriptors do not require coarse quantization of pixel attributes. Instead, in kernel descriptor framework, each pixel equally participates in matching between two image patches. In this paper, we propose a set of kernel descriptors that are based on Tamura features. Additionally, the proposed descriptors are invariant to local rotations. Experimental results show that our proposed approach outperforms the conventional Tamura features significantly.
Integrating object ontology and region semantic template for crime scene investigation image retrieval
- Authors: Liu, Ying , Huang, Yuan , Zhang, Shuai , Zhang, Dengsheng , Ling, Nam
- Date: 2017
- Type: Text , Conference proceedings
- Relation: 2017 12th IEEE Conference on Industrial Electronics and Applications (ICIEA); Siem Reap, Cambodia; 18th-20th June 2017 p. 149-153
- Full Text: false
- Reviewed:
- Description: Crime Scene Investigation (CSI) image retrieval plays an important role in solving crimes by providing useful clues for the police force. However, there has been little work done in this area due to limited public data access by researchers. Tested on real-world CSI images, it was observed that existing content-based image retrieval (CBIR) methods do not necessarily retrieve as effectively on CSI image database as compared to other general image databases. Hence, it is important to design CBIR algorithm tuned to CSI image database. This paper proposes a region-based semantic learning method based on object ontology which associates image categories with 'objects' in CSI images. Each object corresponds to a pre-defined semantic template (ST) which is defined as the average color and texture feature of a set of sample regions. In this way, low-level features of each region in a CSI image can be converted to an 'object' by comparing the region features with the set of pre-defined STs. The 'objects' in an image categorize the image based on the object ontology. The above process is referred to as 'On-Set'. To further improve retrieval performance of On-Set, a weighting strategy named object-frequency-based weighting (OFW) is designed inspired by the idea of term frequency-inverse document frequency (TF-IDF). In OFW, heavier weight is assigned to regions that appear more often in one class and less often in other classes. Experimental results on real-world image data proved the effectiveness of the proposed method for CSI image database retrieval.
Multi-feature fusion for Crime Scene Investigation image retrieval
- Authors: Liu, Ying , Hu, Dan , Fan, Jiulun , Wang, Fuping , Zhang, Dengsheng
- Date: 2017
- Type: Text , Conference proceedings
- Relation: 2017 International Conference on Digital Image Computing : Techniques and Applications (DICTA); Sydney, Australia; 29th November-1st December 2017 p. 865-871
- Full Text: false
- Reviewed:
- Description: Based on a large scale crime scene investigation (CSI) image database, an effective and efficient CSI image retrieval system has been proposed to empower the investigative work of the police force. The main contribution of this paper includes: (1) a DCT domain texture feature extraction algorithm is proposed for CSI images, which is shown to be simple and effective. (2) the use of GIST descriptor on CSI images for the first time and combined with color histogram and the DCT domain texture feature as a fused feature, which describes CSI images from different aspects including color, texture, and scene content. Experimental results prove that the proposed method is effective for CSI image retrieval.
Combining pyramid match kernel and spatial pyramid for image classification
- Authors: Karmakar, Priyabrata , Teng, Shyh , Zhang, Dengsheng , Lu, Guojun , Liu, Ying
- Date: 2016
- Type: Text
- Relation: 2016 International Conference on Digital Image Computing: Techniques and Applications (Dicta); Gold Coast, Australia; 30th November-2nd December 2016 p. 486-493
- Full Text: false
- Reviewed:
- Description: This paper proposes a new approach for image classification by combining pyramid match kernel (PMK) with spatial pyramid. Unlike the conventional spatial pyramid matching (SPM) approach which only uses a single-resolution feature vector to represent an image, we use a multi-resolution feature vector to represent an image for SPM. We then calculate the match scores at each resolution of SPM representation and finally compute the matching between two images by applying the concept of PMK using the match scores obtained from the multiple resolutions. Our experimental results show that the proposed combined pyramid matching achieves a significant improvement on classification performance.
Digital image retrieval using intermediate semantic features and multistep search
- Authors: Zhang, Dengsheng , Liu, Ying , Hou, Jin
- Date: 2008
- Type: Text , Conference paper
- Relation: Proceedings of the Digital Image Computing: Techniques and Applications p. 513-518
- Full Text: false
- Reviewed:
- Description: Recently, semantic image retrieval has attracted large amount of interest due to the rapid growth of digital image storage. However, existing approaches have severe limitations. In this paper, a new approach to digital image retrieval using intermediate semantic features and multistep search has been proposed. Instead of looking for human level semantics which is too challenging at this stage, the research looks for heuristic information and intermediate semantic features which can describe image content objectively. Different from the conventional approaches, the intermediate features are used as filters to eliminate large amount of irrelevant images. Conventional content based image retrieval techniques and relevance feedback (RF) are applied following the filtering to improve the retrieval accuracy. The proposed system has the power of capturing both regional features and global features, and making use of both semantic features and low level features. The proposed system also uses a powerful user interface to provide users with convenient retrieval mechanisms including SQL, RF and query by example. Results show the system has a significant gain over existing region based and global image retrieval approaches
Region-based image retrieval with high-level semantics using decision tree learning
- Authors: Liu, Ying , Zhang, Dengsheng , Lu, Guojun
- Date: 2008
- Type: Text , Journal article
- Relation: Pattern Recognition Vol. 41, no. 8 (2008), p. 2554-2570
- Full Text: false
- Reviewed:
- Description: Semantic-based image retrieval has attracted great interest in recent years. This paper proposes a region-based image retrieval system with high-level semantic learning. The key features of the system are: (1) it supports both query by keyword and query by region of interest. The system segments an image into different regions and extracts low-level features of each region. From these features, high-level concepts are obtained using a proposed decision tree-based learning algorithm named DT-ST. During retrieval, a set of images whose semantic concept matches the query is returned. Experiments on a standard real-world image database confirm that the proposed system significantly improves the retrieval performance, compared with a conventional content-based image retrieval system. (2) The proposed decision tree induction method DT-ST for image semantic learning is different from other decision tree induction algorithms in that it makes use of the semantic templates to discretize continuous-valued region features and avoids the difficult image feature discretization problem. Furthermore, it introduces a hybrid tree simplification method to handle the noise and tree fragmentation problems, thereby improving the classification performance of the tree. Experimental results indicate that DT-ST outperforms two well-established decision tree induction algorithms ID3 and C4.5 in image semantic learning.