Information
Registration Number
0220U104340, 0117U006081 , R & D reports
Title
Intellectual processing of graphic information
popup.stage_title
Head
Ausheva Nataliia M.,
Registration Date
16-12-2020
Organization
National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute"
popup.description2
The report contains 88 sheets. Typewriting text, 27 fig., 74 bibliographic sources. For res. works: 1. In res. experiment. it was proposed to construct a surface on the basis of a conformal substitution of a parameter in the equation of a fifth-order isotropic Bézier curve. The curve was. based on the deformation of a flat isotropic curve. With the help of coefficients. osn. of quadratic shapes it was shown that the surface will be minimal with orthogonal and isothermal grid. Further research is related to the use of set functions. variable in quasi-conformal substitution of parameters. 2. On the basis. lane research developed a program. system thickening of input signals specified as time dependences. Calculate experiments prov. for four interpolation methods: Lagrange, ordinary interpolation Gaussian function, param. Gaussian function, total Gaussian function on approx. elementary algebras. functions and disease statistics. COVID-19 on the territory. Ukraine and Italy. 3. Achieved less. interpolation errors on the main. Gaussian functions by changing the variable parameter α. Cf. with standard pairs. α and with modified. Res. executive research, substantiated the benefits of using. Gaussian methods of interpolation on algebras. functions and real extras. data. 4. On the basis. lane analysis of existing methods of enhancements. point classifications on convolutional networks of basic architectures are determined, which are the most effective methods: architecture (ScanNet), textbook. TreeNet model ensemble, integration of several backbone networks (CBNet). 5. To calculate. experiments implemented the following modif. base. convolution network architectures: ScanNet, TreeNet, CBNet and their combinations: ScanNet + TreeNet, ScanNet + CBNet. On the basis. lane calculated exp. led effect. Mark. methods compared to databases. arch. 6. The method of modif. deep learning networks on the basis. coercion. vizn. contours by the Sobel operator to increase the pixel accuracy
Product Description
popup.authors
Ausheva Nataliia M
Baranichenko Oleksii M
Horbenko Oleksii Yu
Humennyi Arkadii A
Moskalenko Yurii V
Sydorenko Iuliia V
Shapovalova Svitlana I
popup.nrat_date
2020-12-16
search.subscribing
Updated: 2025-12-07
