This document discusses a new multiscale parameterization and history matching workflow for structured and unstructured grids. At the regional scale, it introduces a novel grid-connectivity-based parameterization that flexibly captures heterogeneity while reducing the number of parameters. At the local scale, it leverages established streamline-based methods to refine preferential flow paths. The workflow iterates between regional and local scales to history match production data in a computationally efficient manner.
Gray Image Watermarking using slant transform - digital image processingNITHIN KALLE PALLY
SLANT TRANSFORM
This section details the proposed semi-blind watermarking scheme for copyright protection of digital images. The following subsections present the steps involved in the watermark embedding and extraction processes along with a brief description about the Slant transform, DWT and SVD
Slant transform is derived from saw tooth waveforms. A slant transform matrix has a constant basis vector corresponding to zero sequency and slant basis vector basis vectors monotonically decreases with sequency inconstant step from maximum to minimum. The matrix has sequency property and high energy compaction property[12]. The lowest order of slant matrix is 2 and 2 X 2 matrix is identical to Hadamard matrix
Initial study and implementation of the convolutional Perfectly Matched Layer...Arthur Weglein
In this report, first steps and results of the implementation of the Convolutional Perfectly
Matched Layer (CPML), for the modeling of the 2D acoustic heterogeneous wave equation
are presented. We also compare the conditions to set to zero, for all angles of incidence, the
reflection coefficient at the interface between two PML media, with the analogous conditions
for the reflection coefficient at an interface between two acoustic media. A side product of the
present work for the M-OSRP is a code to create synthetic data, using Finite-Difference (FD)
methods with PML BCs.
We also provide a short description of the main stages involved in the original Reverse Time
Migration (RTM) algorithm, with focus on the 2D acoustic heterogeneous wave equation. We
include a derivation of the equations of the CPML for the backward propagation of the data,
which is part of the RTM. As far as the authors knowledge, these equations and derivations
have not been reported in the literature. The reason we include the RTM is because the present
report can be considered part of a broader research project whose objective is to compare the
RTM with PML BCs with the Green’s theorem based RTM, developed within the M-OSRP.
Gray Image Watermarking using slant transform - digital image processingNITHIN KALLE PALLY
SLANT TRANSFORM
This section details the proposed semi-blind watermarking scheme for copyright protection of digital images. The following subsections present the steps involved in the watermark embedding and extraction processes along with a brief description about the Slant transform, DWT and SVD
Slant transform is derived from saw tooth waveforms. A slant transform matrix has a constant basis vector corresponding to zero sequency and slant basis vector basis vectors monotonically decreases with sequency inconstant step from maximum to minimum. The matrix has sequency property and high energy compaction property[12]. The lowest order of slant matrix is 2 and 2 X 2 matrix is identical to Hadamard matrix
Initial study and implementation of the convolutional Perfectly Matched Layer...Arthur Weglein
In this report, first steps and results of the implementation of the Convolutional Perfectly
Matched Layer (CPML), for the modeling of the 2D acoustic heterogeneous wave equation
are presented. We also compare the conditions to set to zero, for all angles of incidence, the
reflection coefficient at the interface between two PML media, with the analogous conditions
for the reflection coefficient at an interface between two acoustic media. A side product of the
present work for the M-OSRP is a code to create synthetic data, using Finite-Difference (FD)
methods with PML BCs.
We also provide a short description of the main stages involved in the original Reverse Time
Migration (RTM) algorithm, with focus on the 2D acoustic heterogeneous wave equation. We
include a derivation of the equations of the CPML for the backward propagation of the data,
which is part of the RTM. As far as the authors knowledge, these equations and derivations
have not been reported in the literature. The reason we include the RTM is because the present
report can be considered part of a broader research project whose objective is to compare the
RTM with PML BCs with the Green’s theorem based RTM, developed within the M-OSRP.
MDCT audio coding with pulse vector quantizersEricsson
This paper describes a novel audio coding algorithm that is a building block in the recently standardized 3GPP EVS codec. The presented scheme operates in the Modified Discrete Cosine Transform (MDCT) domain and deploys a Split-PVQ pulse coding quantizer, a noise-fill, and a gain control optimized for the quantizer’s properties. A complexity analysis in terms of WMOPS is presented to illustrate that the proposed Split-PVQ concept and dynamic range optimized MPVQ-indexing are suitable for real-time audio coding.
#PR12 #PR366
안녕하세요 논문 읽기 모임 PR-12의 366번째 논문리뷰입니다.
올해가 AlexNet이 나온지 10주년이 되는 해네요.
AlexNet이 2012년에 혜성처럼 등장한 이후, Solve computer vision problem = Use CNN이 공식처럼 사용되던 2010년대가 가고
2020년대 들어서 ViT의 등장을 시작으로 Transformer 기반의 network들이 CNN의 자리를 위협하고 상당부분 이미 뺏어간 상황입니다.
2020년대에 CNN의 가야할 길은 어디일까요?
Inductive bias가 적은 Transformer가 대용량의 데이터로 학습하면 항상 CNN보다 더 낫다는 건 진실일까요?
이 논문에서는 2020년대를 위한 CNN이라는 제목으로 ConvNeXt라는 새로운(?) architecture를 제안합니다.
사실 새로운 건 없고 그동안 있었던 것들과 Transformer에서 적용한 것들을 copy해와서 CNN에 적용해보았는데요,
Transformer보다 성능도 좋고 속도도 빠른 결과가 나왔다고 합니다.
결과에 대해서 약간의 논란이 twitter 상에서 나오고 있는데 이 부분 포함해서 자세한 내용은 영상을 통해서 보실 수 있습니다.
늘 재밌게 봐주시고 좋아요 댓글 구독 해주시는 분들께 감사드립니다 :)
논문링크: https://arxiv.org/abs/2201.03545
영상링크: https://youtu.be/Mw7IhO2uBGc
ConvNeXt: A ConvNet for the 2020s explainedSushant Gautam
Explained here: https://youtu.be/aBvDPL1jFnI
In Nepali
A ConvNet for the 2020s (Zhuang Liu et al.)
ComvNeXt paper
Deep Learning for Visual Intelligence
Sushant Gautam
MSCIISE
Department of Electronics and Computer Engineering
Institute of Engineering, Thapathali Campus
13 March 2022
To all the authors (obviously!!)
1. Jinwon Lee's slides at https://www.slideshare.net/JinwonLee9/pr366-a-convnet-for-2020s?qid=274bc524-23ae-4c13-b03b-0d2416976ad5&v=&b=&from_search=1
2. Letitia from AI Coffee Break: https://www.youtube.com/watch?v=SndHALawoag
I even edited some of her hard visual works and put them as a slide. :(
The use of Orfeo Toolbox in the context of map updatingmelaneum
The use of Orfeo Toolbox in the context of map updating
Christophe Simler; Royal Military Academy
Charles Beumier; Royal Military Academy
Christine Leignel; Université Libre de Bruxelles
Olivier Debeir; Université Libre de Bruxelles
Eléonore Wolff; Université Libre de Bruxelles
The 3GPP evolution for the 3G mobile system created the new base station system, called Evolved UMTS Terrestrial Radio Access Network (E-UTRAN) and a new core network, called Evolved Packet Core (EPC) as a result of two standardisation projects: Long Term Evolution (LTE) and System Architecture Evolution (SAE). Under these specifications a mobile phone gets access to higher bandwidth with low latency in an improved and more efficient network architecture. The standards define an all-IP network as a base for the E-UTRAN/EPC. The E-UTRAN/EPC does not have a separate PS data traffic and CS voice network, both communicate over the same new Evolved Packet System (EPS) network. LTE/EPS Technology course is an intermediate technical course, which covers all aspects of architecture and functionality of the EPS.
The efficient interleaving of digital-video-broadcasting-satellite 2nd genera...TELKOMNIKA JOURNAL
The DVB-S2 system is designed as a toolbox to permit the execution of the satellite programs. Interleaver is an essential part of the DVB-S2 system. The current general block interleaver in DVB-S2 is not best, which leads to high BER and maybe not satisfy the requirements of the system. The purpose of this paper is to study the several interleaver types and comparative analyses are done between them to find which of these give better performance. Simulations results obtained prove that the 2D interleavers minimize BER more than other interleavers of DVB-S2. Further, the performance of 2D interleaver is better on a system that required a low SNR.
International Journal of Computational Engineering Research (IJCER) is dedicated to protecting personal information and will make every reasonable effort to handle collected information appropriately. All information collected, as well as related requests, will be handled as carefully and efficiently as possible in accordance with IJCER standards for integrity and objectivity.
SVD and Lifting Wavelet Based Fragile Image WatermarkingIDES Editor
Creation and distribution of digital multimedia, by
copying and editing, has both advantages and disadvantages.
These can facilitate unauthorized usage, misappropriation,
and misrepresentation. Therefore the content providers have
become more concerned. So image watermarking, which is the
act of embedding another signal (the watermark) into an
image, have been proposed for copyright protection and
authentication by robust and fragile methodologies
respectively. So for various applications, there are different
watermarking algorithms, but here this work is mainly for
authentication as the watermarking scheme is fragile. The
discrete lifting based wavelet transform and the singular value
decomposition (SVD) algorithms are used in this scheme. The
former for the carrier or the image to be authenticated, while
the latter for the logo which is embedded in the carrier. The
distribution of SVD compressed pixel values are distributed in
the wavelet domain based on a pseudorandom sequence. This
has been observed to test the integrity of the stego image and
its authentication. Moreover due to usage of lifting based
wavelet transform and SVD the hardware implementability is
better.
Usability of GI – case study using address locationsMuki Haklay
Nick Bearman's presentation in the 2nd Workdshop on usability of geographic information, 23rd March 2010 at UCL, London. See details at http://www.virart.nottingham.ac.uk/GI%20Usability/index.html
This paper aims, a 3D-Pilot Aided Multi-Input Multi-Output Orthogonal Frequency Division Multiplexing (MIMO-OFDM) Channel Estimation (CE) for Digital Video Broadcasting -T2 (DVB-T2) for the 5 different proposed block and comb pilot patterns model and performed on different antenna configuration. The effects of multi-transceiver antenna on channel estimation are addressed with different pilot position in frequency, time and the vertical direction of spatial domain framing. This paper first focus on designing of 5-different proposed spatial correlated pilot pattern model with optimization of pilot overhead. Then it demonstrates the performance comparison of Least Square (LS) & Linear Minimum Mean Square Error (LMMSE), two linear channel estimators for 3D-Pilot Aided patterns on different antenna configurations in terms of Bit Error Rate. The simulation results are shown for Rayleigh fading noise channel environments. Also, 3x4 MIMO configuration is recommended as the most suitable configuration in this noise channel environments.
Modification on Energy Efficient Design of DVB-T2 Constellation De-mapperIJERA Editor
The second generation of terrestrial digital video broadcasting standard (DVB-T2) offers several advantages for greater efficiency. Signal Space Diversity (SSD) contains rotated constellation and Q-Delay (RQD), which is one of advantage that offered to improve the performance over fading channels compared to the non-rotated modulation. In this journal, the proposed low-power de-mapper design of this work attempts to employ the introduced SSD to reduce power through replacing LLR calculations by a significantly less complex projection-based de-mapping whenever possible. It benefits from an algorithm that applies projection-based de-mapping to significantly reduce LLR computations without deteriorating performance. Two versions are introduced for hard de-mapping and soft de-mapping. The design uses several techniques simultaneously to be even more energy efficient without affecting the performance. Prototype results indicate significant reduction of LLR calculations as Eb/N0 increases with no performance degradation. The idea and energy saving techniques can be easily applied to any rotated constellation de-mapper.
Data - the Oil & Gas asset that isn’t managed like oneMolten2013
This is Molten's view on data management in the Oi l& Gas that Rory Colfer, Managing Partner, has recently presented at the ECIM E&P Data and IM Conference in Europe. For more information or to read on "How oil and gas executives could use ‘Big Data’ as a powerful source of competitive advantage", please visit our website http://www.molten-group.com/the-data-crunch/
MDCT audio coding with pulse vector quantizersEricsson
This paper describes a novel audio coding algorithm that is a building block in the recently standardized 3GPP EVS codec. The presented scheme operates in the Modified Discrete Cosine Transform (MDCT) domain and deploys a Split-PVQ pulse coding quantizer, a noise-fill, and a gain control optimized for the quantizer’s properties. A complexity analysis in terms of WMOPS is presented to illustrate that the proposed Split-PVQ concept and dynamic range optimized MPVQ-indexing are suitable for real-time audio coding.
#PR12 #PR366
안녕하세요 논문 읽기 모임 PR-12의 366번째 논문리뷰입니다.
올해가 AlexNet이 나온지 10주년이 되는 해네요.
AlexNet이 2012년에 혜성처럼 등장한 이후, Solve computer vision problem = Use CNN이 공식처럼 사용되던 2010년대가 가고
2020년대 들어서 ViT의 등장을 시작으로 Transformer 기반의 network들이 CNN의 자리를 위협하고 상당부분 이미 뺏어간 상황입니다.
2020년대에 CNN의 가야할 길은 어디일까요?
Inductive bias가 적은 Transformer가 대용량의 데이터로 학습하면 항상 CNN보다 더 낫다는 건 진실일까요?
이 논문에서는 2020년대를 위한 CNN이라는 제목으로 ConvNeXt라는 새로운(?) architecture를 제안합니다.
사실 새로운 건 없고 그동안 있었던 것들과 Transformer에서 적용한 것들을 copy해와서 CNN에 적용해보았는데요,
Transformer보다 성능도 좋고 속도도 빠른 결과가 나왔다고 합니다.
결과에 대해서 약간의 논란이 twitter 상에서 나오고 있는데 이 부분 포함해서 자세한 내용은 영상을 통해서 보실 수 있습니다.
늘 재밌게 봐주시고 좋아요 댓글 구독 해주시는 분들께 감사드립니다 :)
논문링크: https://arxiv.org/abs/2201.03545
영상링크: https://youtu.be/Mw7IhO2uBGc
ConvNeXt: A ConvNet for the 2020s explainedSushant Gautam
Explained here: https://youtu.be/aBvDPL1jFnI
In Nepali
A ConvNet for the 2020s (Zhuang Liu et al.)
ComvNeXt paper
Deep Learning for Visual Intelligence
Sushant Gautam
MSCIISE
Department of Electronics and Computer Engineering
Institute of Engineering, Thapathali Campus
13 March 2022
To all the authors (obviously!!)
1. Jinwon Lee's slides at https://www.slideshare.net/JinwonLee9/pr366-a-convnet-for-2020s?qid=274bc524-23ae-4c13-b03b-0d2416976ad5&v=&b=&from_search=1
2. Letitia from AI Coffee Break: https://www.youtube.com/watch?v=SndHALawoag
I even edited some of her hard visual works and put them as a slide. :(
The use of Orfeo Toolbox in the context of map updatingmelaneum
The use of Orfeo Toolbox in the context of map updating
Christophe Simler; Royal Military Academy
Charles Beumier; Royal Military Academy
Christine Leignel; Université Libre de Bruxelles
Olivier Debeir; Université Libre de Bruxelles
Eléonore Wolff; Université Libre de Bruxelles
The 3GPP evolution for the 3G mobile system created the new base station system, called Evolved UMTS Terrestrial Radio Access Network (E-UTRAN) and a new core network, called Evolved Packet Core (EPC) as a result of two standardisation projects: Long Term Evolution (LTE) and System Architecture Evolution (SAE). Under these specifications a mobile phone gets access to higher bandwidth with low latency in an improved and more efficient network architecture. The standards define an all-IP network as a base for the E-UTRAN/EPC. The E-UTRAN/EPC does not have a separate PS data traffic and CS voice network, both communicate over the same new Evolved Packet System (EPS) network. LTE/EPS Technology course is an intermediate technical course, which covers all aspects of architecture and functionality of the EPS.
The efficient interleaving of digital-video-broadcasting-satellite 2nd genera...TELKOMNIKA JOURNAL
The DVB-S2 system is designed as a toolbox to permit the execution of the satellite programs. Interleaver is an essential part of the DVB-S2 system. The current general block interleaver in DVB-S2 is not best, which leads to high BER and maybe not satisfy the requirements of the system. The purpose of this paper is to study the several interleaver types and comparative analyses are done between them to find which of these give better performance. Simulations results obtained prove that the 2D interleavers minimize BER more than other interleavers of DVB-S2. Further, the performance of 2D interleaver is better on a system that required a low SNR.
International Journal of Computational Engineering Research (IJCER) is dedicated to protecting personal information and will make every reasonable effort to handle collected information appropriately. All information collected, as well as related requests, will be handled as carefully and efficiently as possible in accordance with IJCER standards for integrity and objectivity.
SVD and Lifting Wavelet Based Fragile Image WatermarkingIDES Editor
Creation and distribution of digital multimedia, by
copying and editing, has both advantages and disadvantages.
These can facilitate unauthorized usage, misappropriation,
and misrepresentation. Therefore the content providers have
become more concerned. So image watermarking, which is the
act of embedding another signal (the watermark) into an
image, have been proposed for copyright protection and
authentication by robust and fragile methodologies
respectively. So for various applications, there are different
watermarking algorithms, but here this work is mainly for
authentication as the watermarking scheme is fragile. The
discrete lifting based wavelet transform and the singular value
decomposition (SVD) algorithms are used in this scheme. The
former for the carrier or the image to be authenticated, while
the latter for the logo which is embedded in the carrier. The
distribution of SVD compressed pixel values are distributed in
the wavelet domain based on a pseudorandom sequence. This
has been observed to test the integrity of the stego image and
its authentication. Moreover due to usage of lifting based
wavelet transform and SVD the hardware implementability is
better.
Usability of GI – case study using address locationsMuki Haklay
Nick Bearman's presentation in the 2nd Workdshop on usability of geographic information, 23rd March 2010 at UCL, London. See details at http://www.virart.nottingham.ac.uk/GI%20Usability/index.html
This paper aims, a 3D-Pilot Aided Multi-Input Multi-Output Orthogonal Frequency Division Multiplexing (MIMO-OFDM) Channel Estimation (CE) for Digital Video Broadcasting -T2 (DVB-T2) for the 5 different proposed block and comb pilot patterns model and performed on different antenna configuration. The effects of multi-transceiver antenna on channel estimation are addressed with different pilot position in frequency, time and the vertical direction of spatial domain framing. This paper first focus on designing of 5-different proposed spatial correlated pilot pattern model with optimization of pilot overhead. Then it demonstrates the performance comparison of Least Square (LS) & Linear Minimum Mean Square Error (LMMSE), two linear channel estimators for 3D-Pilot Aided patterns on different antenna configurations in terms of Bit Error Rate. The simulation results are shown for Rayleigh fading noise channel environments. Also, 3x4 MIMO configuration is recommended as the most suitable configuration in this noise channel environments.
Modification on Energy Efficient Design of DVB-T2 Constellation De-mapperIJERA Editor
The second generation of terrestrial digital video broadcasting standard (DVB-T2) offers several advantages for greater efficiency. Signal Space Diversity (SSD) contains rotated constellation and Q-Delay (RQD), which is one of advantage that offered to improve the performance over fading channels compared to the non-rotated modulation. In this journal, the proposed low-power de-mapper design of this work attempts to employ the introduced SSD to reduce power through replacing LLR calculations by a significantly less complex projection-based de-mapping whenever possible. It benefits from an algorithm that applies projection-based de-mapping to significantly reduce LLR computations without deteriorating performance. Two versions are introduced for hard de-mapping and soft de-mapping. The design uses several techniques simultaneously to be even more energy efficient without affecting the performance. Prototype results indicate significant reduction of LLR calculations as Eb/N0 increases with no performance degradation. The idea and energy saving techniques can be easily applied to any rotated constellation de-mapper.
Data - the Oil & Gas asset that isn’t managed like oneMolten2013
This is Molten's view on data management in the Oi l& Gas that Rory Colfer, Managing Partner, has recently presented at the ECIM E&P Data and IM Conference in Europe. For more information or to read on "How oil and gas executives could use ‘Big Data’ as a powerful source of competitive advantage", please visit our website http://www.molten-group.com/the-data-crunch/
User guide of reservoir geological modeling v2.2.0Bo Sun
This is the user guide of DepthInsight™ reservoir geological modeling module. For corresponding video tutorials , please visit and subscribe our Youtube channel: https://www.youtube.com/channel/UCjHyG-mG7NQofUWTZgpBT2w
DepthInsight™ software products include modules as follows:
Structure Interpretation
Well and Data Management
Plan Module
Profile Module
Attribute Modeling
Velocity Modeling
Structural Modeling
Reservoir Geological Modeling
Numerical Simulation Gridding
Rock Modeling
Geo-mechanical Modeling
Paleo-Structural Modeling
Enormous Modeling Platform
For more information about our company, Beijing GridWorld Software Technology Co., Ltd., please visit our website: http://gridworld.com.cn/en/
Introduction Petrel Course (UAB-2014)
This course has been prepared as an introduction of Petrel software (Schlumberger, www.software.slb.com/products/platform/Pages/petrel.aspx), an application which allows the modeling and visualization of reservoirs, since the exploration stage until production, integrating geological and geophysical data, geological modeling (structural and stratigraphic frameworks), well planning, or property modeling ( petrophysical or petrological) among other possibilities.
The course will be focused mainly in the understanding and utilization of workflows aimed to build geological models based on superficial data (at the outcrop scale) but also with seismic data. The course contents have been subdivided in 5 modules each one developed through the combination of short explanations and practical exercises.
The duration of the course covers more or less 10h divided in three sessions. The starting data will be in the first week of December.
This course will be oriented mainly for the PhD and master students ascribed at the Geologic department of the UAB. For logistic reasons the maximum number of places for each torn are 9. The course is free from the Department members but the external interested will have to make a symbolic payment.
Those interested send an e-mail to the Doctor Griera (albert.griera@uab.cat).
The course will be imparted by Marc Diviu (Msc. Geology and Geophysics of reservoirs).
3D Facies Modelling project using Petrel software. Msc Geology and Geophysics
Abstract
The Montserrat and Sant Llorenç del Munt fan-delta complexes were developed during the Eocene in the Ebro basin. The depositional stratigraphic record of these fan deltas has been described as a made up by a several transgressive and regressive composite sequences each made up by several fundamental sequences. Each sequence set is in turn composed by five main facies belts: proximal alluvial fan, distal alluvial fan, delta front, carbonates platforms and prodelta.
Using outcrop data from three composite sequences (Sant Vicenç, Vilomara and Manresa), a 3D facies model was built. The key sequential traces of the studied area georeferenced and digitalized on to photorealistic terrain models, were the hard data used as input to reconstruct the main surfaces, which are separating transgressive and regressive stacking patterns. Regarding the facies modelling has been achieved using a geostatistical algorithm in order to define the stacking trend and the interfingerings of adjacent facies belts, and five paleogeographyc maps to reproduce the paleogeometry of the facies belts within each system tract.
The final model has been checked, using a real cross section, and analysed in order to obtain information about the Delta Front facies which are the ones susceptible to be analogous of a reservoir. Attending to the results including eight probability maps of occurrence, the transgressive sequence set of Vilomara is the greatest accumulation of these facies explained by its agradational component.
Optimization of distributed generation of renewable energy sources by intelli...Beniamino Murgante
Optimization of distributed generation of renewable energy sources by intelligent techniques
Marcello Pucci – Institute for Studies on Intelligent Systems for Automation (I.S.S.I.A), National Research Council, Palermo (Italy)
Intelligent Analysis of Environmental Data (S4 ENVISA Workshop 2009)
Optimization of distributed generation of renewable energy sources by intelli...Beniamino Murgante
Optimization of distributed generation of renewable energy sources by intelligent techniques
Marcello Pucci – Institute for Studies on Intelligent Systems for Automation (I.S.S.I.A), National Research Council, Palermo (Italy)
Intelligent Analysis of Environmental Data (S4 ENVISA Workshop 2009)
Optimization of distributed generation of renewable energy sources by intelli...Beniamino Murgante
Optimization of distributed generation of renewable energy sources by intelligent techniques
Marcello Pucci – Institute for Studies on Intelligent Systems for Automation (I.S.S.I.A), National Research Council, Palermo (Italy)
Intelligent Analysis of Environmental Data (S4 ENVISA Workshop 2009)
InternImage: Exploring Large-Scale Vision Foundation Models with Deformable C...taeseon ryu
요즘 대형 비전 트랜스포머(ViT)의 발전에 비해, 합성곱 신경망(CNN)을 기반으로 한 대형 모델은 아직 초기 단계에 머물러 있습니다. 본 연구는 InternImage라는 새로운 대규모 CNN 기반 모델을 제안합니다. 이 모델은 ViT와 같이 매개변수와 학습 데이터를 늘리는 이점을 얻을 수 있습니다. 최근에는 대형 밀집 커널에 초점을 맞춘 CNN과는 달리, InternImage는 변형 가능한 컨볼루션을 핵심 연산자로 사용합니다. 이를 통해 모델은 감지 및 세분화와 같은 하향 작업에 필요한 큰 유효 수용영역을 갖게 되며, 입력 및 작업 정보에 의존하는 적응형 공간 집계도 가능합니다. 이로 인해, InternImage는 기존 CNN의 엄격한 귀납적 편향을 줄이고, ViT와 같은 대규모 매개변수와 대규모 데이터로 더 강력하고 견고한 패턴을 학습할 수 있게 됩니다. 논문에서 제시한 모델의 효과성은 ImageNet, COCO 및 ADE20K와 같은 어려운 벤치마크에서 입증되었습니다. InternImage-H는 COCO test-dev에서 65.4 mAP, ADE20K에서 62.9 mIoU를 달성하여 현재 최고의 CNN 및 ViT를 능가하는 새로운 기록을 세웠습니다
Comprehending Web Applications by a Clustering Based Approach Porfirio Tramontana
The number and the complexity of web applications are increasing dramatically to satisfy the market requests, and the need of effective approaches for comprehending them is growing accordingly. Recently, some reverse engineering methods and tools have been proposed to support the comprehension of a web application; the information recovered by these tools is usually rendered in graphical representations. However, the graphical representations become progressively less useful with large-scale applications, and do not support adequately the comprehension of the application.
In this paper, to overcome this limitation, we propose an approach based on a clustering method for decomposing a web application (WA) into groups of highly functionally related components. The approach is based on the definition of a coupling measure between interconnected components of the WA that takes into account both the typology and the topology of the connections. The coupling measure is exploited by a clustering algorithm that produces a hierarchy of clustering. This hierarchy allows a structured approach to the comprehension of the web application to be carried out. The approach has been experimented with on medium sized web applications and produced interesting and encouraging results.
Vision Transformer(ViT) / An Image is Worth 16*16 Words: Transformers for Ima...changedaeoh
computer vision 분야에서 dominant 한 Convolutional Layer를 일절 사용하지 않고, NLP에서 제안된 순수 Transformer의 architecture를 그대로 가져와 Attention과 일반 Feed Forward NN만을 이용하여 SOTA수준의 Image Classification Model을 구축한다.
TAVE research seminar 21.03.30 발표자료
발표자: 오창대
Bhark, E.W., Structured History Matching Workflow using Parameterization and Streamline Methods
1. Multiscale Parameterization and
History Matching in Structured and
Unstructured Grids:
Theory and Field Application
E. W. Bhark, A. Rey, A. Datta-Gupta and B. Jafarpour
2. Motivation
• Develop structured history matching workflow
• Coarse (regional) scale
Novel grid-connectivity-based
parameterization
• Flexible, efficient application for
large models, complex geology
Calibrate multiscale heterogeneity
Avoid traditional regional multipliers
• Local (grid cell) scale
Established streamline-based method
• Vasco et al. (1998); Datta-Gupta and King (2007)
Refine prior preferential flow paths
2
3. Outline of presentation
• Parameterization in history matching
Methods of linear transformation
Grid-connectivity-based parameterization
• Structured history matching workflow
• Field application
Offshore reservoir model (Rey et al. [2009], SPE124950)
3
4. Why re-parameterization?
• Reduce redundant model information
Preserve important heterogeneity
~5,000 Unknowns 100 Unknowns 50 25
Ex., high-resolution
(3D) abs. permeability
• Improves:
Solution non-uniqueness and stability, computational efficiency
4
5. Parameterization by linear transform
= v1 + v2 + v3 + … + v50 + … + vN
N-parameter
high-resolution
model
u= v Φ M N u v for M << N
u1
1
u
2 v1
• Required basis properties
2 v2
Compression power: most
energy in fewest coefficients vi
M
v M
Amenable to efficient
u
N application for very large grids
5
6. Highlights of new basis
u1
1 u
2 v
2 v
Grid-connectivity-based transform basis =
M
v M
(1) Model (or prior) independent u
N
Can benefit from prior model information
(2) Applicable to any grid geometry (e.g., CPG, irregular unstructured,
NNCs, faults)
(3) Efficient construction for very large grids
(4) Strong, generic compression performance
(5) Geologic spatial continuity
6
7. Basis development
Concept: Develop as generalization of discrete Fourier basis
KEY: Perform Fourier transform of function u by (scalar) projection
on eigenvectors of grid Laplacian (2nd difference matrix)
• Interior rows
Second difference
Periodic operator (circulant matrix)
• Exterior rows
Boundary conditions control
eigenvector behavior
7
8. Basis development
CPG Unstructured Grid Laplacian
5
10
15
20
25
30
35
40
45
50
5 10 15 20 25 30 35 40 45 50
2-point connectivity (1/2/3-D)
• Decompose L to construct basis functions (rows of )
Always symmetric, sparse
Efficient (partial) decomposition by restarted Lanczos method
Orthogonal basis functions; Φu v u Φ1 v ΦT v
• In general (non-periodic) case
Eigen(Lanczos)vectors vibrational modes of the model grid
Eigenvalues represent modal frequencies
8
10. Basis functions: Examples
Unstructured grid Basis function 1 Basis function 3 Basis function 5 Basis function 8 Basis function 10
Unstructured grid
(local refinement)
Channel structure
Multiple subdomains
10
11. Structured multiscale workflow
(1) START: Prior model (2) Regional update (3) Local update
Prior spatial hydraulic Parameterize
property model Streamline-,
multiplier field
sensitivity-based
inversion (GTTI)
Update in transform
domain
Multiscale iterate
Gradient-based
iterate
Back-transform
Unit-multiplier field at multiplier field to
grid cell resolution spatial domain
Calibrated Model
FINISH
Flow and transport
Add higher-
simulation
frequency modes to
basis
NO Data misfit
tolerance?
YES
Additional YES
spatial
detail?
NO
11
12. Field application: Offshore reservoir
Reservoir
• > 300,000 cells
• Mature waterflood
• 8 years of production history
• 4 producers and 4 water injectors
• Complex depositional sequence of
turbidite sand bodies / facies
• Rey et al. (2009), SPE124950
Parameter
• Permeability
Data
• Water cut
12
13. Conceptual heterogeneity model
Prior model facies (5)
Prior geo-model P2
I2
P1
P3 I1
Initial Kx: I3
Average of measurements
at wells per facies (5) Facies ID
P4
I4
Next objective:
Use parameterization to assist
in heterogeneity identification
and updating
13
14. Workflow: Prior model & multiplier field
F2
Prior geo-model
Multiplier field
F6
F5
F3
F1
14
15. Facies basis functions
Facies 5:
Prior geo-model • Multiplier field is linear
combination of basis functions
Multiplier field
Basis functions F5 multiplier field:
u=
1 3 6 8 15
v1 …+ v3 …+ v6 …+ v8 …+ v15
15
16. Adaptive multiscale inversion
Prior geo-model • Sequentially refine within-facies heterogeneity
From coarse to finer scales
Adaptive inclusion of basis functions
Multiplier field
1 5 10
Basis functions
Multiscale inversion
• End refinement when production data become
insensitive to addition of basis functions
16
17. Multiscale update
Number of leading basis
Kx: Adaptive multiscale
functions per facies
10
10
10
1
5
36
17
18. Comparison with previous calibration
This study Rey et al. (2009)
Tx multiplier Facies zonation Tx multiplier
Adaptive multiscale Manual zonation
18
23. Comparison of data misfit: WCT
Multiscale/SL and Business Unit
P2 P3
P4 P1
23
24. Comparison with previous calibration
P3
This study • Regional
I3 I4
SOURCE parameterization
I2
more consistent with
model constraints
I3
I4
TMX: Rey et al. (2009) Figure 26: Rey et al. (2009)
TMX
mult.
High perm
(> upper limit
near P3)
Potential
channel
24
25. Summary
• Multiscale approach to history matching
Builds on well-established ‘structured’ workflow
Regional heterogeneity
Generalized grid-connectivity-based parameterization
Efficient, flexible application to any reservoir model geometry
Refine local heterogeneity
Prior preferential flow paths captured by streamlines
• Field application
Demonstrates practical feasibility
Improvement upon heterogeneity characterization using
standard zonation approaches
25