• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.10
• /
• pp.2643-2657
• /
• 2023

Scene graphs are structured representations that can clearly convey objects and the relationships between them, but are often heavily biased due to the highly skewed, long-tailed relational labeling in the dataset. Indeed, the visual world itself and its descriptions are biased. Therefore, Unbiased Scene Graph Generation (USGG) prefers to train models to eliminate long-tail effects as much as possible, rather than altering the dataset directly. To this end, we propose Geometric and Semantic Improvement (GSI) for USGG to mitigate this issue. First, to fully exploit the feature information in the images, geometric dimension and semantic dimension enhancement modules are designed. The geometric module is designed from the perspective that the position information between neighboring object pairs will affect each other, which can improve the recall rate of the overall relationship in the dataset. The semantic module further processes the embedded word vector, which can enhance the acquisition of semantic information. Then, to improve the recall rate of the tail data, the Class Balanced Seesaw Loss (CBSLoss) is designed for the tail data. The recall rate of the prediction is improved by penalizing the body or tail relations that are judged incorrectly in the dataset. The experimental findings demonstrate that the GSI method performs better than mainstream models in terms of the mean Recall@K (mR@K) metric in three tasks. The long-tailed imbalance in the Visual Genome 150 (VG150) dataset is addressed better using the GSI method than by most of the existing methods.

• Korean Journal of Computational Design and Engineering
• /
• v.1 no.2
• /
• pp.97-106
• /
• 1996

Existing CAD systems do not provide advanced functions for automatic checking design and drafting errors in mechanical drawings. If the knowledge of checking in mechanical ddrsfting can be implemented into computers, CAD systems could automatically check for design and drafting errors. This paper describes a method for systematic checking of dimension errors. such as deficiency and/or redundancy of dimension input-errors in dimension figures and symbols, etc. The logic for finding dimensional errors is written by using a proccedural language. A geometric model and a topological-graph model are used in this method. Checking for deficiency and redundancy of dimensions is based upon graph Theory.

• Journal of the Korean Data and Information Science Society
• /
• v.9 no.1
• /
• pp.63-70
• /
• 1998

The singular value decomposition is a tool which is used to find a linear structure of reduced dimension and to give interpretation of the lower dimensional structure about multivariate data. In this paper the singular value decomposition is reviewed from both algebraic and geometric point of view and, is illustrated the way which the tool is used in the multivariate techniques finding a simpler geometric structure for the data.

• Journal of applied mathematics & informatics
• /
• v.5 no.1
• /
• pp.13-22
• /
• 1998

Fractals which represent many of the sets in various scien-tific fields as well as in nature is geometrically too complicate. Then we usually use Hausdorff dimension to estimate their geometrical proper-ties. But to explain the fractals from the hausdorff dimension induced by the Euclidan metric are not too sufficient. For example in digi-tal communication while encoding or decoding the fractal images we must consider not only their geometric sizes but also many other fac-tors such as colours densities and energies etc. So in this paper we define the dimension matrix of the sets by redefining the new metric.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.19 no.3
• /
• pp.418-424
• /
• 1994

Like the Hermitian function field over GF(q), those subfields defined by y +y=x where s divides q+1 are also maximal, having the maximum number os places of degree one permissible by the Hasse-Weil bound. Geometric Goppa codes(or algebraic geometric codes) arising from these subfields of the Hermitian function field are studied in this paper. Their dimension and minimum distance are explicilty and completely presented for any m with m

• The Research Journal of the Costume Culture
• /
• v.12 no.1
• /
• pp.59-72
• /
• 2004

The objectives of this study were to investigate the perceptions of Korean female university students for clothing pattern tendency and structural element of clothing pattern image dimension and to find how individual personality type influence the preferred clothing pattern characteristics. For this study, a questionnaire was designed and sent to 600 female university students of Daejeon, Seoul and metropolitan area. The tool used in this study was MBTI(The Myers-Briggs Type Indicator) Form G Korean version and for the analysis of data SPSS 10.0 package were used. 10 representative patterns for this study were floral, dot, stripe, check, animal, abstract & artistic, geometric, vegetable & leaf, paisely, patchwork pattern. The data were analyzed by factor analysis, arithmetic mean, One-Way ANOVA, x²-test. The major findings were as follows: Clothing pattern image dimension perceived by Korean female university students for 10 representative patterns were basic form, deluxe, specialty, and cultural dimension. Among them, basic form and deluxe dimension were major dimensions. In basic form dimension, dot pattern score was high indicating female students perceive it as light, comfortable, clean, cool and simple pattern image. In deluxe dimension, floral pattern scored high and in specialty dimension, abstract and artistic pattern scored high among other pattern image. In cultural dimension, geometric pattern and check pattern scored high. Based on other detailed analysis results, It is concluded that the personality type greatly influence clothing pattern evaluation. For example, in case of color combination of patchwork pattern, there was a difference in color preference depend on a personality type such as sensing(S) or intuition(N). Therefore, sensing personality type preferred adjacent color combination than contrast color combination. Detailed marketing strategy is necessary in planning textile design of merchandise plan.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.34S no.11
• /
• pp.114-125
• /
• 1997

This paper presents a method that automatically recognizes dimension sets from the mechanical drawings, and that classifies 6 types dimension sets according to functional purpose. In the proposed method, the object and closed-loop symbols are separated from the character-free drawings. Then object lines and interpretation lines are vectorized. And, after recognizing dimension sets(consistings of arrowhead, shape line, tail lines, extension lines, text-string, and feature control frame), we classify recognized dimension sets as horizontal, vertical, angular, diametral, radial, and leader dimension sets. Finally the proposed method converts classified dimension sets into AutoCAD data by using AutoLisp language. By using the methods of geometric modeling, the proposed method readily recognized and classifies dimension sets from complex drawings. Experimetnal results are presented, which are obtained by applying the proposed method to drawings drawn in compliance with the KS drafting standard.

• Journal of Institute of Control, Robotics and Systems
• /
• v.22 no.9
• /
• pp.710-715
• /
• 2016

Due to its simplicity and high reliability, the box-counting(BC) method is one of the most frequently used techniques to estimate the fractal dimensions of a binary image with a self-similarity property. The fractal calculation requires data sampling that determines the size of boxes to be sampled from the given image and directly affects the accuracy of the fractal dimension estimation. There are three non-overlapping regular grid methods: geometric-step method, arithmetic-step method and divisor-step method. These methods have some drawbacks when the image size M becomes large. This paper presents a BC algorithm for enhancing the accuracy of the fractal dimension estimation based on a new sampling method. Instead of using the geometric-step method, the new sampling method, called the coverage ratio-step method, selects the number of steps according to the coverage ratio. A set of experiments using well-known fractal images showed that the proposed method outperforms the existing BC method and the triangular BC method.

• Communications for Statistical Applications and Methods
• /
• v.6 no.1
• /
• pp.261-266
• /
• 1999

We present a geometric interpretation of Chebyshev type inequalities. This uses a simple diagram which illustrates the functional bound for the indicator function of the event whose probability we want to assess. We also give a geometric interpretation of the inequalities in terms of volume in a Euclidean space of appropriate dimension. Markov's inequality and Chebyshev's inequality are treated in more detail.

• Structural Engineering and Mechanics
• /
• v.13 no.2
• /
• pp.231-239
• /
• 2002

The problem of dynamic analysis of truss structures based on probability is studied in this paper. Considering the randomness of both physical parameters (elastic module and mass density) of structural materials and geometric dimension of bars respectively or simultaneously, the stiffness and mass matrixes of the elements and structure have been built. The structure dynamic characteristic based on probability is analyzed, and the expressions of numeral characteristics of inherence frequency random variable are derived from the Rayleigh's quotient. The method of structural dynamic analysis based on probability is developed. Finally, two examples are given.