• Title/Summary/Keyword: floating-point image

Search Result 41, Processing Time 0.027 seconds

A low-cost compensated approximate multiplier for Bfloat16 data processing on convolutional neural network inference

  • Kim, HyunJin
    • ETRI Journal
    • /
    • v.43 no.4
    • /
    • pp.684-693
    • /
    • 2021
  • This paper presents a low-cost two-stage approximate multiplier for bfloat16 (brain floating-point) data processing. For cost-efficient approximate multiplication, the first stage implements Mitchell's algorithm that performs the approximate multiplication using only two adders. The second stage adopts the exact multiplication to compensate for the error from the first stage by multiplying error terms and adding its truncated result to the final output. In our design, the low-cost multiplications in both stages can reduce hardware costs significantly and provide low relative errors by compensating for the error from the first stage. We apply our approximate multiplier to the convolutional neural network (CNN) inferences, which shows small accuracy drops with well-known pre-trained models for the ImageNet database. Therefore, our design allows low-cost CNN inference systems with high test accuracy.

Formative Expressions by Artificial Light applied to Office Building Lobbies (현대 오피스 로비공간에서 빛의 조형적 표현 특성에 관한 연구)

  • Jeong, Soo-Ryun
    • Korean Institute of Interior Design Journal
    • /
    • v.18 no.2
    • /
    • pp.41-49
    • /
    • 2009
  • Contemporary design environment is formed with image-centered trend based on pluralism. In this point of view, enterprises' building lobbies are public places containing the equivocal meaning, actively utilizing light as a design element to express the image of enterprises' identifications. Light is an immaterial entity having unlimited possibilities and potentials on space. It also acts as media to activate spaces and create new images in connection with formative elements of space. This study is to figure out how lightings are expressed and affected the formative characteristics of office lobby spaces and activate the specific characteristics of spaces. As a result, we drew conclusions as follows. First, as state-of-the-art technology and media are introduced, light is expressed on spaces as floating, direction, rhythm, silhouette, metaphor and allusion, sense of depth and volume. Second, expressive aspects of light in lobby space are embodiment of light, substantiation of immateriality, standing of materiality from the perspective of spatial aesthetics, and distortion/transformation of shape, pluralism phenomena of space from the perspective of spatial structure. In this way, light on building lobbies which are greatly required design differentiation strategy, specializes space and also integrates all the designs as not only a functional element but also a mental, psychological, formative element. Consequently, light on lobby spaces induces communication between spaces and users, makes formative value of existence in itself, and presents the characteristics of differentiated enterprises' identities.

A Study on the Dematerializing Tendency in Contemporary Space Design (공간디자인의 비(非)물질화 경향에 관한 연구)

  • 권영걸
    • Korean Institute of Interior Design Journal
    • /
    • no.22
    • /
    • pp.61-68
    • /
    • 2000
  • This paper illustrates what the reality of architectural concepts, materials, notions and phenomena are as central questions of contemporay space design. These issues form the goal of modern space design which should attain to the reality of an era when non-real values prevail. Despite the trends of an everchanging and ephemeral dominating quality in architecture for the last ten years, architects still aim to contruct everlasting space on earth. The trends of dematerialization in today's space design can be substantiated in spatial-temporal dimensions as follows. First of all, ephemeral architecture with concepts of hypothetical temporality, everchanging architecture in fluidity, and the transparent architecture with the floating and overlapping image can be analyzed in the dimension of 'time'. In terms of 'space', void quality for the expression of emptiness, neutral space by the simplified and summarized forms, expanded space through ambiguous boundaries and spatial repetition can becharacterized and also be intended strategies for lightness, state of flux, ambiguity, paradox etc., lead modern space design along that path. As this point, we need to pay attention to the so-called 'hypersurface' concept proposed by Stephen Perella. Hypersurface is a sort of cladding sheathing existence independent from the primary structure. With it, the integration between form and image can be achieved. Sometimes hypersurface can be a strategical screen for image projection, a cognitive receptor for surroundings as well as a catalyst for information and communication systems. When the situation dematerializes more and more as the years go on, the concept of hypersurface can be an inclusive method between the phenomenological form in architecture and its self recipient image. Permissive atmospheres created between them in contemporary space design and new paradigms emerged with digital technology will further reinforce the human space's dynamism.

  • PDF

Design of Special Function Unit for Vectorized SIMD Programmable Unified Shader (벡터화된 SIMD 프로그램어블 통합 셰이더를 위한 특수 함수 유닛 설계)

  • Jung, Jin-Ha;Kim, Kyeong-Seob;Yun, Jeong-Hee;Seo, Jang-Won;Choi, Sang-Bang
    • Journal of the Institute of Electronics Engineers of Korea SD
    • /
    • v.47 no.5
    • /
    • pp.56-70
    • /
    • 2010
  • Rendering technique generating 2 dimensional image to give reality and high performance graphical processor for efficient processing of massive data are necessary to support realistic 3 dimensional graphical image. Recently, graphical hardwares have evolved rapidly. This enables high quality rendering effect that we were unable to process in realtime. Improving shading technique enabled us to render realistic images but still much time is required for this process. Multiple operational units are being integrated in a graphical processor for effective floating point operation using massive data to process almost real looking images. In this paper, we have designed and implemented a special functional unit to support high quality 3 dimensional computer graphic image on programmable integrated shader processor. We have done evaluation through functional level simulation of designed special functional unit. Hardware resource usage rate and execution speed are measured implementing directly on FPGA Virtex-4(xc4vlx200).

Float Zone Growth of Superconduction $Bi_2Sr_2CaCu_2O_8$ Fiber with Halogen Lamp (할로겐 램프 열원을 이용한 초전도성 $Bi_2Sr_2CaCu_2O_8$ Fiber의 Float Zone 성장)

  • 김철진;정준기
    • Korean Journal of Crystallography
    • /
    • v.7 no.2
    • /
    • pp.147-155
    • /
    • 1996
  • Floating Zone Image furnace with halogen lamps as heat source has been made and applied to the growth of high-Tc superconductor Bi2Sr2CaCu2O8. The manufactured crystal growth equipment is composed of holder unit for setting the halogen lamps at the focal point of the elliptical mirror, image furnace with maximum temperature of 1800℃ with 1kW halogen lamps, cooling unit, feeding unit for supplying source material to the molten zone, pulling unit for crystal growth, and the control unit in the range of 2mm/hr -40 mm/hr vertical movement and 15rpm - 12rpm rotation. Bi2Sr2CaCu2O8 fibers have been grown with 300W halogen lamps and characterized with XRD, SEM and EDS. The growth condition was air atmosphere, growth speed 3∼4mm/hr, rotation speed of upper and lower part 20∼25 rpm. The fiber was composed of 20∼25 rpm matrix phase and secondary phases such as (Sr,Ca)CuO2 and (Sr,Ca)2CuO3.

  • PDF

Digital painting: Image transfonnation, simulation, heterologie and transfonnation (현대회화에서의 형태와 물질 -Digital Transfiguration에 관한 연구-)

  • Jeong, Suk-Yeong
    • Journal of Science of Art and Design
    • /
    • v.10
    • /
    • pp.161-181
    • /
    • 2006
  • The words which appeared in my theoretical study and work are image transformation to digital painting, simulation, heterologie and transfiguration, etc. Firstly, let's look into 'digital era' or 'new media era'. Nowadays, the image world including painting within the rapid social and cultural change, which is called as digital era, is having the dramatic change. Together with the development of scientific technology, large number of events which was deemed to be impossible is happening as real in image world Moreover, these changes in image world is greatly influencing to our life. The word which compresses this change of image world and shows is 'digital'. Digit, which means fingers in Latin, indicates separately changing signal, and to be more narrow, it indicates the continual signal of '0' and ' 1' in computer. The opposite word is 'analogue'. As analogue is the word meaning 'infer' or 'similarity', it indicates the signal or form which continuously changes along the series of time when it is compared to digital. Instead of analogue, digital is embossed as a major ruler along the whole area of our current culture. In whole culture and art area, and in whole generalscience, digital is appearing as it has the modernism and importance. The prefix, 'digital', e.g. digital media, digital culture, digital design, digital philosophy, etc, is treated as the synonym of modernism and something new. This advent of digital results the innovative change to the image world, creates the new beauty experience which we could not experience before, and forecasts the formation of advanced art and expansion of creative area. Various intellectual activities using computer is developing the whole world with making the infrastructure. Computer in painting work immediately accomplishes the idea of painters, takes part in simulation work, contingency such as abrupt reversal, extraction, twisting, shaking, obscureness, overlapping, etc, and timing to stimulate the creativity of painters, and provides digital formative language which enables new visual experience to the audience. When the change of digital era, the image appeared in my work is shown in 'transfiguration' like drawing. The word, 'transfiguration' does not indicate the completed and fixed real substance but indicate endlessly moving and floating shape. Thus, this concept is opposite to the substantial consideration, so that various concepts which is able to replace this in accordance with the similar cases are also exist such as change, deterioration, mutation, deformity of appearance and morphing which is frequently used in computer as a technical word. These concepts are not clearly classified, and variably and complicatedly related. Transfiguration basically means the denial of "objectivity' and '(continual) stagnation' or deviation from those. This phenomenon is appeared through the all art schools of art ever since the realism is denied in the 19th century. It is called as 'deformation' in case of expressionism, futurism, cubism, etc, in the beginning of the century, which its former indication is mostly preserved within the process of structural deviation and which has the realistic limit which should be preserved. On the contrary, dramatic transfiguration which has been showing in the modern era through surrealism is different in the point that dramatic transfiguration tends to show the deterioration and deviation rather than the preservation of indicated object. From this point, transfiguration coming out from morphing using computer deteriorates and hides the reality and furthermore, it replaces the 'reality'. Moreover, transfiguration is closely approached to the world of fake or 'imaginary' simulation world of Baudrillard. According to Baudrillard, the image hides and deteriorates the reality, and furthermore, expresses 'not existing' to 'imaginary' under the name of transfiguration. Certain reality, that is, image which is absent from the reality is created and overflowed, so that it finally replaces the reality. This is simulation as it is said by Baudrillard. In turn, Georges Bataille discusses about the image which is produced by digital technology in terms of heterologie. Image of heterologie is the visual signal which is established with the media. Image of media is to have the continuous characteristics of produce, extinction, and transformation, and its clear boundary between images becomes meaningless. The meaning of composition, excess, violation, etc of digital image is explained to heterological study or heteologie suggested as important meaning of Georges Bataille who is a heretic philosopher. As the form and image of mutation shows the shape in accordance with mechanical production, heterologie is introduced as very low materialism (or bas materialisme), in this theory. Heterologie as low materialism which is gradually changing is developing as a different concept and analysis because of the change of time in the late 20s century beside high or low meaning. Including my image, all images non-standardizes and transforms the code. However, reappearance and non-standardization of this code does not seem to be simple. The problem of transformation caused by transfiguration which appears in my digital drawing painting, simulation, heterologie, etc, are the continual problems. Moreover, the subject such as existence of human being, distance from the real life, politics and social problems are being extended to actual research and various expressing work. Especially, individual image world is established by digital painting transfiguration technique, and its change and review start to have the durability. The consciousness of observers who look at the image is changing the subject. Together with theoretical research, researchers are to establish the first step to approach to various image change of digital era painting through transfiguration technique using our realistic and historical image.

  • PDF

A VLSI Architecture of an 8$\times$8 OICT for HDTV Application (HDTU용 8$\times$8 최적화 정수형 여현 변환의 VLSE 구조)

  • 송인준;황상문;이종하;류기수;곽훈성
    • Journal of the Korean Institute of Telematics and Electronics T
    • /
    • v.36T no.1
    • /
    • pp.1-7
    • /
    • 1999
  • We present VLSI architecture for a high performance 2-D DCT processor which is used compressing system of real time image processing or HDTV using fast computational algorithm of the Optimized Integer Cosine Transform(OICT). The coefficients of the OICT are integer, so the OICT performs only the integer operations for both forward and inverse transform. Therefore the proposed architecture could be greatly enhanced in improving the speed, reduced the hardware cost considerably by replacing the multiplication operations with shift and addition operations compared with DCT which performs floating-point operations.

  • PDF

RLDB: Robust Local Difference Binary Descriptor with Integrated Learning-based Optimization

  • Sun, Huitao;Li, Muguo
    • KSII Transactions on Internet and Information Systems (TIIS)
    • /
    • v.12 no.9
    • /
    • pp.4429-4447
    • /
    • 2018
  • Local binary descriptors are well-suited for many real-time and/or large-scale computer vision applications, while their low computational complexity is usually accompanied by the limitation of performance. In this paper, we propose a new optimization framework, RLDB (Robust-LDB), to improve a typical region-based binary descriptor LDB (local difference binary) and maintain its computational simplicity. RLDB extends the multi-feature strategy of LDB and applies a more complete region-comparing configuration. A cascade bit selection method is utilized to select the more representative patterns from massive comparison pairs and an online learning strategy further optimizes descriptor for each specific patch separately. They both incorporate LDP (linear discriminant projections) principle to jointly guarantee the robustness and distinctiveness of the features from various scales. Experimental results demonstrate that this integrated learning framework significantly enhances LDB. The improved descriptor achieves a performance comparable to floating-point descriptors on many benchmarks and retains a high computing speed similar to most binary descriptors, which better satisfies the demands of applications.

Quality Analysis on Computer Generated Hologram Depending on the Precision on Diffraction Computation (회절연산 정밀도에 따른 CGH 기반 홀로그램 생성 품질 분석)

  • Jaehong Lee;Duksu Kim
    • Journal of Broadcast Engineering
    • /
    • v.28 no.1
    • /
    • pp.21-30
    • /
    • 2023
  • Computer-generated holography requires much more computation costs and memory space rather than image processing. We implemented the diffraction calculation with low-precision and mixed-precision floating point numbers and compared the processing time and quality of the hologram with various precision. We compared diffraction quality with double, single and bfloat16 precision. bfloat16 shows 5.94x and 1.52x times faster performance than double precision and single precision. Also, bfloat16 shows lower PSNR and SSIM and higher MSE than other precision. However, there is no significant effect on reconstructed images. These results show low precision, like bfloat16, can be utilized for computer-generated holography.

Comparison of 3D Space Perception for the Stereoscopic AR Holography (스테레오 증강현실 홀로그래피에서의 삼차원 공간감 비교)

  • Kim, Minju;Wohn, Kwangyun
    • Journal of the HCI Society of Korea
    • /
    • v.8 no.2
    • /
    • pp.21-27
    • /
    • 2013
  • Recently, the use of floating hologram has increased in many different aspects, such as exhibitions, education, advertisements, and so on. Especially, the floating hologram that makes use of half-mirror is widely used. Nevertheless, half-mirror, unfortunately, cannot lead users to the perfect three dimensional hologram experience. Even though it can make the vision look to be up on the air, it does not have the capacity to display itself up on the air, which is the ultimate goal of hologram. In addition, it looks inconsistent when a real object is located behind the half-mirror in order to show the convergence of the two (object and the half-mirror). In this paper, we did the study on comparison of 3D space perception for the stereoscopic AR holography. At first, we applied stereoscopic technology to the half-mirror hologram system for the accurate and realistic AR environment. Then, the users can feel as if the real 3D object behind half-mirror and the reflected virtual image are converged much better in the 3D space. Furthermore, by using depth camera, the location and direction of graphics can be controlled to change depending on the user's point of view. This is the effective way to produce augmented stereoscopic images simply and accurately through half-mirror film without any additional devices. What we saw from the user test were applying 3D images and user interaction leads the users to have 3D spatial awareness and realism more effectively and accurately.

  • PDF