Image region annotation based on segmentation and semantic correlation analysis

Zhang, Jing; Mu, Yakun; Feng, Shengwei; Li, Kehuang; Yuan, Yubo; Lee, Chin‐Hui

doi:10.1049/iet-ipr.2017.0917

Cited by 14 publications

(14 citation statements)

References 43 publications

(58 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Semantic ambiguity can be resolved by identifying the various regions in image where multiple labels assigned, which is the motivation to introduced tri-relational graph. In tri-relational graph annotation method, image is divided in different region and set of various region T is prepared [26,31]. Set of semantic labels C and Image set X is prepared.…”

Section: Traditional Graph Verses Tri-relational Graph Learningmentioning

confidence: 99%

“…To generate Image graph, visual similarity of all the regions from are calculated and compared. Looking towards the concept the segmentation of the image is important task in TG which is achieved through Texture-enhanced JSEG algorithm which is depend on regional latent semantic dependency [31]. For correct relatively independent segmentation, texture and colour class map are combined by texture-enhanced segmentation (TJSEG).…”

Section: Fig2mentioning

confidence: 99%

See 1 more Smart Citation

Survey and Analysis on Machine Learning Approaches for Image Annotation

V.R¹,

L²

2021

JUSST

View full text Add to dashboard Cite

In current years, a large amount of image data is being collected worldwide, which is majorly generated by corporate organizations, health industry and social networking sites. With the strength of substantial level depiction of images, Annotating image has numerous applications not only in image understanding and analysis but also in some of the concern domain like medical research, rural and urban management. Automatic Image Annotation (AIA) has been raised since the late 1990s due to inherent weaknesses of manual image annotation. In this paper, a deep review of the most recent stage in the development of AIA methods is presented by synthesizing 32 literatures published during the past decades. We classify AIA methods into ﬁve categories: 1) Kernel Logistic Regression (KLR), 2) Tri-relational Graph (TG), 3) Semantically Regularised CNN- RNN (S-CNN-RNN), 4) Label Correlation guided Deep Multi-view (LCDM), and 5) Multi-Modal Semantic Hash Learning (MMSHL). Considering inspiration on the basis of main idea, framework of model, complexity of computation, time complexity and accuracy in annotation Comparative analysis for various AIA methods are done.

show abstract

Section: Traditional Graph Verses Tri-relational Graph Learningmentioning

confidence: 99%

Section: Fig2mentioning

confidence: 99%

Survey and Analysis on Machine Learning Approaches for Image Annotation

V.R¹,

L²

2021

JUSST

View full text Add to dashboard Cite

show abstract

“…Authors in [54] consider undirected graphical models that jointly exploit low-level features and contextual information (as concept co-occurrences and spatial correlation statistics) to classify local image blocks into predefined concepts. Zhang et al [56] propose a region annotation framework that exploits the semantic correlation of segmented image regions; this method assigns each segmented region to one concept and learns the relationships between labels and region locations using PSA. A hybrid annotation approach based on visual attention mechanism and conditional random fields is proposed in [57] in order to pay more attention to the salient regions during the annotation process.…”

Section: Context Modelingmentioning

confidence: 99%

Deep Context-Aware Kernel Networks

Jiu¹,

Sahbi²

2019

Preprint

View full text Add to dashboard Cite

Context plays a crucial role in visual recognition as it provides complementary clues for different learning tasks including image classification and annotation. As the performances of these tasks are currently reaching a plateau, any extra knowledge, including context, should be leveraged in order to seek significant leaps in these performances. In the particular scenario of kernel machines, context-aware kernel design aims at learning positive semi-definite similarity functions which return high values not only when data share similar contents, but also similar structures (a.k.a contexts). However, the use of context in kernel design has not been fully explored; indeed, context in these solutions is handcrafted instead of being learned. In this paper, we introduce a novel deep network architecture that learns context in kernel design. This architecture is fully determined by the solution of an objective function mixing a content term that captures the intrinsic similarity between data, a context criterion which models their structure and a regularization term that helps designing smooth kernel network representations. The solution of this objective function defines a particular deep network architecture whose parameters correspond to different variants of learned contexts including layerwise, stationary and classwise; larger values of these parameters correspond to the most influencing contextual relationships between data. Extensive experiments conducted on the challenging ImageCLEF Photo Annotation and Corel5k benchmarks show that our deep context networks are highly effective for image classification and the learned contexts further enhance the performance of image annotation.

show abstract

“…Image annotation is one of the major challenges in computer vision which aims at assigning keywords (a.k.a labels or concepts) to images. The difficulty in image annotation stems from the extreme variability of the learned concepts and their versatile content which is usually described with handcrafted or learned representations [1][2][3][4][5][6][7][8][9][10][11]. However, due to its limited representational power, content is usually upgraded with context in order to capture both the intrinsic and the extrin-This work was supported by a grant from the National Natural Science Foundation of China (No.…”

Section: Introductionmentioning

confidence: 99%

DHCN: Deep Hierarchical Context Networks For Image Annotation

Jiu¹,

Sahbi²

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Context modeling is one of the most fertile sub-fields of visual recognition which aims at designing discriminant image representations while incorporating their intrinsic and extrinsic relationships. However, the potential of context modeling is currently under-explored and most of the existing solutions are either context-free or restricted to simple handcrafted geometric relationships. We introduce in this paper DHCN: a novel Deep Hierarchical Context Network that leverages different sources of contexts including geometric and semantic relationships. The proposed method is based on the minimization of an objective function mixing a fidelity term, a context criterion and a regularizer. The solution of this objective function defines the architecture of a bi-level hierarchical context network; the first level of this network captures scene geometry while the second one corresponds to semantic relationships. We solve this representation learning problem by training its underlying deep network whose parameters correspond to the most influencing bi-level contextual relationships and we evaluate its performances on image annotation using the challenging ImageCLEF benchmark.

show abstract

Image region annotation based on segmentation and semantic correlation analysis

Cited by 14 publications

References 43 publications

Survey and Analysis on Machine Learning Approaches for Image Annotation

Survey and Analysis on Machine Learning Approaches for Image Annotation

Deep Context-Aware Kernel Networks

DHCN: Deep Hierarchical Context Networks For Image Annotation

Contact Info

Product

Resources

About