TOWARDS SLAM-AWARE SCENE UNDERSTANDING

Størrelse: px
Begynne med side:

Download "TOWARDS SLAM-AWARE SCENE UNDERSTANDING"

Transkript

1 TOWARDS SLAM-AWARE SCENE UNDERSTANDING Bowl Bowl Co ee Mug Soda Can Cap Sudeep Pillai Marine Robotics Group CSAIL, MIT Jan 27, 2016

2 MOTIVATION Our goal: To enable a robot to capture and understand the spatial structure and semantic properties of their immediate environment 2

3 MOTIVATION SLAM-AWARE RECOGNTION MAP REPRESENTATION SLAM-AWARE LEARNING 2 3 Bowl Bowl Co ee Mug 4 Soda Can Cap 1 M Monocular SLAM Supported Object Recognition S. Pillai & J. Leonard (RSS 15) High-Performance and Tunable Stereo Reconstruction S. Pillai, S. Ramalingam & J. Leonard (ICRA 16) Large-scale SLAM-aware Object Learning (in the works) 3

4 MOTIVATION Robots equipped with a single RGB camera need to continuously recognize and localize all potential objects in its immediate environment Single RGB Camera Versatile Robust Avoid spurious detection/mis-classification Multi-view Object Detection Camera & Object localization by leveraging SLAM Scalable Runtime sub-linear in categories identifiable Input RGB Video 4

5 RECENT WORK Leverage SLAM capabilities Semi-dense reconstructions could potentially propose objects Object Detection / Recognition can be better informed with SLAM ORB-SLAM Mur-Artal et al. LSD-SLAM Engel et al. (ECCV 2014) Detection-based Object Labeling in 3D scenes Lai et al. (ICRA 2012) SLAM++ Salas-Moreno et al. (CVPR 2013) 5 Semi-Dense Mapping with ORB-SLAM Mur-Artal et al. (RSS 2015)

6 SLAM-AWARE RECOGNITION 1 M SLAM-aware {, M} Keyframes Map M 6

7 SLAM-AWARE RECOGNITION 2 1 M SLAM-aware {, M} Keyframes Map M 7

8 SLAM-AWARE RECOGNITION M SLAM-aware {, M} Keyframes Map M 8

9 SLAM-AWARE RECOGNITION M SLAM-aware {, M} Keyframes Map M 9

10 SLAM-AWARE RECOGNITION M Keyframes SLAM-aware {, M} Map M ŷ MLE = Most likely argmax y2{1,..., C } p(d o y) 8 o 2 O semantic label 10

11 KEY CONCEPT SLAM-capable robots equipped with a single RGB camera need to continuously recognize and localize all potential objects in its immediate environment Single RGB Camera (Monocular SLAM supports improved recognition) Robust (Reduced false positives via view correspondence from SLAM) Multi-view Object Detection (Proposals from Semi-Dense Maps) Scalable (FLAIR encoding) SLAM-Supported Object Recognition 11

12 MONOCULAR SLAM-SUPPORTED OBJECT RECOGNITION SEMI-DENSE MAPPING Input RGB Video Scale-ambiguous Reconstruction UW-RGBD Dataset (v2) Lai et al. (ICRA 2014) (a) Semi-Dense Depth Filtering SVO: Forster et al. (ICRA 2014) + ORB-SLAM Mur-Artal et al. As a camera-equipped robot moves around in its immediate environment, the camera is localized and simultaneously a semi-dense map is constructed 12

13 MONOCULAR SLAM-SUPPORTED OBJECT RECOGNITION MULTI-VIEW OBJECT PROPOSALS Scale-ambiguous Reconstruction Filtered Reconstruction (a) Semi-Dense Depth Filtering SVO: Forster et al. (ICRA 2014) + ORB-SLAM Mur-Artal et al. (b) Low-density region pruning The semi-dense reconstruction provide spatio-temporally consistent edges, a strong indication of object presence 13

14 MONOCULAR SLAM-SUPPORTED OBJECT RECOGNITION MULTI-VIEW OBJECT PROPOSALS Scale-ambiguous Reconstruction Multi-scale Over-segmentation (a) Semi-Dense Depth Filtering SVO: Forster et al. (ICRA 2014) + ORB-SLAM Mur-Artal et al. (b) Density-based segmentation over 4 spatial scales Object proposals are extracted from the reconstructed scene via a multi-scale density segmentation step, and are further refined for occlusions 14

15 MONOCULAR SLAM-SUPPORTED OBJECT RECOGNITION MULTI-VIEW OBJECT PROPOSALS Multi-scale Over-segmentation View-Consistent Proposals View View (d) View (b) Density-based segmentation over 4 spatial scales View Candidate proposals projection The segmented regions are projected onto each keyframe enabling view-consistent object proposals that perform better than frame-based proposal methods 15

16 MONOCULAR SLAM-SUPPORTED OBJECT RECOGNITION VIEW AGGREGATION Unified Object Prediction SLAM-Supported Object Recognition Bowl View Bowl Co ee Mug Soda Can Cap Fusion of object proposal classification The object proposals in each frame are encoded (via VLAD + FLAIR or R-CNN) and classified, before their evidence is probabilistically fused across all keyframes 16

17 SLAM-SUPPORTED vs. FRAME-BASED OBJECT RECOGNITION SLAM-Supported Recognition (Ours) Frame-based Recognition (Classical approach) CORRECT PREDICTIONS 17 INCORRECT PREDICTIONS

18 SLAM-SUPPORTED vs. FRAME-BASED OBJECT RECOGNITION Object hypotheses are aggregated across views resulting in correct classification SLAM-Supported Recognition (Ours) Frame-based Recognition (Classical approach) CORRECT PREDICTIONS 18 INCORRECT PREDICTIONS

19 SLAM-SUPPORTED vs. FRAME-BASED OBJECT RECOGNITION 19

20 SLAM-SUPPORTED vs. FRAME-BASED OBJECT RECOGNITION PERFORMANCE Recognition performance RGB: Superior to frame-based methods RGB-D: Comparable to state-of-the-art SLAMbased recognition methods Higher is better SINGLE-VIEW RECOGNITION VLAD + FLAIR encoding: Run-time nearly independent of object categories identifiable Ours (RGB) map DetOnly (RGB) Conclusion SLAM-aware / Multi-view reasoning Object proposals via semi-dense reconstructions Scalable recognition (FLAIR encoding) Higher is better MULTI-VIEW RECOGNITION Comparable results to RGB-D recognition 20 map Ours Det3DMRF HMP2D+3D (RGB) (RGB-D) (RGB-D) map - (mean Average Precision)

21 HIGH-PERFORMANCE AND TUNABLE STEREO RECONSTRUCTION S. Pillai, S. Ramalingam & J. Leonard (ICRA 16) 21

22 MOTIVATION Full Scene Reconstruction using Stereo SGBM / ELAS (~ ms) Scene Discretization and Fusion using Octomap (~ 20 ms) 22

23 MOTIVATION NAVIGATION PLANS (SLOW) Discretized Scene Reconstruction (~ 200 ms / 5 Hz) 23

24 MOTIVATION? Navigation-focused / Plan-Aware Scene 5 m/s (11 m/s (22 20 m/s (45 mph) 5 Hz / FPS 5 m/s => 1 m / frame 10 m/s => 2 m / frame 20 m/s => 4 m / frame 100 Hz / FPS 5 m/s => 0.05 m / frame 10 m/s => 0.1 m / frame 20 m/s => 0.2 m / frame

25 PIECE-WISE PLANAR APPROXIMATION VIA TESSELLATION Step 1 & 2: Establish tessellation and piece-wise planar approximation given the sparse key points and their corresponding disparities 25

26 RECURSIVE DEPTH REFINEMENT Step 3 & 4: Determine incorrectly approximated regions in the image (matching cost check), resample, re-tessellate, and re-estimate stereo disparity in these regions 26

27 RECURSIVE DEPTH REFINEMENT VIA TESSELLATION Fast piece-wise planar approximation Extremely efficient ( Hz) 1 Iteration Iterative refinement (any-time performance) Tunable (high-speed/low-accuracy or highaccuracy/low-speed regimes) 2 Iterations Surprisingly accurate for reconstruction Depth prior 4 Iterations 2 iterations: 90 % within 3px disparity 80 Hz 27

28 FAST YET SURPRISINGLY POWERFUL Depth prior determined via Delaunay triangulation of sparse support points. Disparities for the sparse support points are obtained via sparse-stereo matching 28

29 SEMI-DENSE DISPARITY ESTIMATION Interpolated semi-dense disparities estimated from the piecewise planar tessellation 29

30

31 CONSISTENT STEREO RECONSTRUCTIONS AT HIGH-SPEEDS High-performance and tunable stereo algorithm Fastest semi-dense stereo algorithm to-date Piece-wise planar approximation Any-time performance Adaptive algorithm that enables for low-speed / high-quality or high-speed / low-quality stereo reconstructions Compelling reconstructions with fused VO for high-speed navigation Locally-consistent stereo reconstructions using our high-performance stereo algorithm coupled with stereo visual odometry 31

32 SLAM-AWARE LEARNING Wide/Narrow Baseline Stereo RGB/Depth Camera Laser Rangefinder Simulated Turtlebot (Gazebo) in an environment

33 SLAM-AWARE LEARNING Simulated Robot Exploration Large-scale data collection Life-long object learning / autonomy Object Detector evolution Objects as landmarks Train in simulation, adapt to real-world Coupled recognition and SLAM Scalable scene description Viewpoint and Lighting Invariant Human interpretable Simulated Turtlebot (Gazebo) exploring in an environment. SLAM provides view correspondences

34 SLAM-AWARE OBJECT LEARNING 2 3 Tag detections are registered using deadreckoned (odometry) pose 1 M 4 Single object instance / Multiple-views Dead-reckoned pose trajectory of robot

35 THANKS! Contact: Sudeep Pillai Webpage: 35

Satellite Stereo Imagery. Synthetic Aperture Radar. Johnson et al., Geosphere (2014)

Satellite Stereo Imagery. Synthetic Aperture Radar. Johnson et al., Geosphere (2014) Satellite Stereo Imagery Synthetic Aperture Radar Johnson et al., Geosphere (2014) Non-regular sampling Missing data due to lack of correlation, shadows, water, Potentially 3D as opposed to purely 2D (i.e.

Detaljer

UNIK 4690 Maskinsyn Introduksjon

UNIK 4690 Maskinsyn Introduksjon UNIK 4690 Maskinsyn Introduksjon 19.01.2017 Trym Vegard Haavardsholm (trymh@ifi.uio.no) Idar Dyrdal (idar@unik.no) Thomas Opsahl (Thomas-Olsvik.Opsahl@ffi.no) Ragnar Smestad (Ragnar.Smestad@ffi.no) Maskinsyn

Detaljer

CAMES. Technical. Skills. Overskrift 27pt i to eller flere linjer teksten vokser opad. Brødtekst 22pt skrives her. Andet niveau.

CAMES. Technical. Skills. Overskrift 27pt i to eller flere linjer teksten vokser opad. Brødtekst 22pt skrives her. Andet niveau. CAMES Overskrift 27pt i to eller flere linjer Technical Skills Leizl Joy Nayahangan, RN, MHCM Leizl.joy.nayahangan@regionh.dk IMPORTANCE Challenges Brødtekst 22pt of patient skrives her care Increasing

Detaljer

Andrew Gendreau, Olga Rosenbaum, Anthony Taylor, Kenneth Wong, Karl Dusen

Andrew Gendreau, Olga Rosenbaum, Anthony Taylor, Kenneth Wong, Karl Dusen Andrew Gendreau, Olga Rosenbaum, Anthony Taylor, Kenneth Wong, Karl Dusen The Process Goal Definition Data Collection Data Preprocessing EDA Choice of Variables Choice of Method(s) Performance Evaluation

Detaljer

Den europeiske byggenæringen blir digital. hva skjer i Europa? Steen Sunesen Oslo,

Den europeiske byggenæringen blir digital. hva skjer i Europa? Steen Sunesen Oslo, Den europeiske byggenæringen blir digital hva skjer i Europa? Steen Sunesen Oslo, 30.04.2019 Agenda 1. 2. CEN-veileder til ISO 19650 del 1 og 2 3. EFCA Guide Oppdragsgivers krav til BIMleveranser og prosess.

Detaljer

Neural Network. Sensors Sorter

Neural Network. Sensors Sorter CSC 302 1.5 Neural Networks Simple Neural Nets for Pattern Recognition 1 Apple-Banana Sorter Neural Network Sensors Sorter Apples Bananas 2 Prototype Vectors Measurement vector p = [shape, texture, weight]

Detaljer

UNIK 4690 Maskinsyn Introduksjon

UNIK 4690 Maskinsyn Introduksjon UNIK 4690 Maskinsyn Introduksjon 21.01.2016 Trym Vegard Haavardsholm (trymh@ifi.uio.no) Idar Dyrdal (idar@unik.no) Thomas Opsahl (Thomas-Olsvik.Opsahl@ffi.no) Ragnar Smestad (Ragnar.Smestad@ffi.no) Maskinsyn

Detaljer

Splitting the differential Riccati equation

Splitting the differential Riccati equation Splitting the differential Riccati equation Tony Stillfjord Numerical Analysis, Lund University Joint work with Eskil Hansen Innsbruck Okt 15, 2014 Outline Splitting methods for evolution equations The

Detaljer

Estimating Peer Similarity using. Yuval Shavitt, Ela Weinsberg, Udi Weinsberg Tel-Aviv University

Estimating Peer Similarity using. Yuval Shavitt, Ela Weinsberg, Udi Weinsberg Tel-Aviv University Estimating Peer Similarity using Distance of Shared Files Yuval Shavitt, Ela Weinsberg, Udi Weinsberg Tel-Aviv University Problem Setting Peer-to-Peer (p2p) networks are used by millions for sharing content

Detaljer

Accuracy of Alternative Baseline Methods

Accuracy of Alternative Baseline Methods Accuracy of Alternative Baseline Methods Dr. Steven Braithwait Christensen Associates Energy Consulting IEPEC - Paris June 2010 Outline Demand response & role of baseline loads Measures of baseline performance

Detaljer

SVM and Complementary Slackness

SVM and Complementary Slackness SVM and Complementary Slackness David Rosenberg New York University February 21, 2017 David Rosenberg (New York University) DS-GA 1003 February 21, 2017 1 / 20 SVM Review: Primal and Dual Formulations

Detaljer

BPS TESTING REPORT. December, 2009

BPS TESTING REPORT. December, 2009 BPS TESTING REPORT December, 2009 Standardized Testing BPS SAT Reasoning Test SAT Subject Tests Advanced Placement ACT Massachusetts Comprehensive Assessment System (MCAS) MCAS Growth Data 2 SAT Reasoning

Detaljer

SENSORS. HAIN An Integrated Acoustic Positioning and Inertial Navigation System

SENSORS. HAIN An Integrated Acoustic Positioning and Inertial Navigation System SENSORS HAIN An Integrated Acoustic Positioning and Inertial Navigation System Hans Petter Jacobsen and Jan Erik Faugstadmo Session Chair Steve Browne, Thales Geo Solutions September 16-17, 2003 Houston,

Detaljer

Dean Zollman, Kansas State University Mojgan Matloob-Haghanikar, Winona State University Sytil Murphy, Shepherd University

Dean Zollman, Kansas State University Mojgan Matloob-Haghanikar, Winona State University Sytil Murphy, Shepherd University Dean Zollman, Kansas State University Mojgan Matloob-Haghanikar, Winona State University Sytil Murphy, Shepherd University Investigating Impact of types of delivery of undergraduate science content courses

Detaljer

2A September 23, 2005 SPECIAL SECTION TO IN BUSINESS LAS VEGAS

2A September 23, 2005 SPECIAL SECTION TO IN BUSINESS LAS VEGAS 2A September 23, 2005 SPECIAL SECTION TO IN BUSINESS LAS VEGAS SPECIAL SECTION TO IN BUSINESS LAS VEGAS 3A September 23, 2005 SEE, PAGE 8A Businesses seek flexibility. It helps them compete in a fast-paced,

Detaljer

Multimedia in Teacher Training (and Education)

Multimedia in Teacher Training (and Education) Multimedia in Teacher Training (and Education) Bodo Eckert, Stefan Altherr, Hans-Jörg Jodl Second International GIREP Seminar 1-6 September 2003 University of Udine, Italy Content Training courses for

Detaljer

Lattice Simulations of Preheating. Gary Felder KITP February 2008

Lattice Simulations of Preheating. Gary Felder KITP February 2008 Lattice Simulations of Preheating Gary Felder KITP February 008 Outline Reheating and Preheating Lattice Simulations Gravity Waves from Preheating Conclusion Reheating and Preheating Reheating is the decay

Detaljer

Dynamic Programming Longest Common Subsequence. Class 27

Dynamic Programming Longest Common Subsequence. Class 27 Dynamic Programming Longest Common Subsequence Class 27 Protein a protein is a complex molecule composed of long single-strand chains of amino acid molecules there are 20 amino acids that make up proteins

Detaljer

OPPA European Social Fund Prague & EU: We invest in your future.

OPPA European Social Fund Prague & EU: We invest in your future. OPPA European Social Fund Prague & EU: We invest in your future. Talk Outline appearance based tracking patch similarity using histogram tracking by mean shift experiments, discussion Mean shift Tomáš

Detaljer

BEC in microgravity. W. Herr, Institute for Quantum Optics, Leibniz University of Hanover

BEC in microgravity. W. Herr, Institute for Quantum Optics, Leibniz University of Hanover BEC in microgravity W. Herr, Institute for Quantum Optics, Leibniz University of Hanover Outline Reminder: Why microgravity? Why BEC? BEC in microgravity: Pilot project, Quantus Results Ongoing activity

Detaljer

Moving Objects. We need to move our objects in 3D space.

Moving Objects. We need to move our objects in 3D space. Transformations Moving Objects We need to move our objects in 3D space. Moving Objects We need to move our objects in 3D space. An object/model (box, car, building, character,... ) is defined in one position

Detaljer

Generalisering med vektor tiles

Generalisering med vektor tiles Generalisering med vektor tiles Bjørn Sandvik MasterMaps Nordisk kartografikurs 2018 Tønsberg 25. - 27. september 2018 MasterMaps.com norviz.com Scale and generalization Selection: choosing which objects

Detaljer

Little Mountain Housing

Little Mountain Housing Little Mountain Housing Feedback from January 2012 Open Houses Presentation to Little Mountain Community Advisory Group Overview Open house attendance 409 signed in 600+ total Comment forms submitted 326

Detaljer

VLSI Design for Yield on Chip Level

VLSI Design for Yield on Chip Level IBM Systems and Technology Group Markus Bühler Jeanne Bickford Jason Hibbeler Jürgen Koehl DATE 2006 Outline Catastrophic Failures Defect Mechanisms State of the Art Novel Techniques Conclusion 2 Catastrophic

Detaljer

Feiltre, hendelsestre og RIF-modell

Feiltre, hendelsestre og RIF-modell Initiating Event BB4 Initiating Event Type 3 End Control Type Type 2 End Control 2 B5/C2 Feiltre, hendelsestre og RIFmodell Rolf Bye, Studio Apertura Initiating Event structure C & C3 Omission structure

Detaljer

Emnedesign for læring: Et systemperspektiv

Emnedesign for læring: Et systemperspektiv 1 Emnedesign for læring: Et systemperspektiv v. professor, dr. philos. Vidar Gynnild Om du ønsker, kan du sette inn navn, tittel på foredraget, o.l. her. 2 In its briefest form, the paradigm that has governed

Detaljer

OPPA European Social Fund Prague & EU: We invest in your future.

OPPA European Social Fund Prague & EU: We invest in your future. OPPA European Social Fund Prague & EU: We invest in your future. Talk Outline appearance based tracking patch similarity using histogram tracking by mean shift experiments, discussion Mean shift Tomáš

Detaljer

GeWare: A data warehouse for gene expression analysis

GeWare: A data warehouse for gene expression analysis GeWare: A data warehouse for gene expression analysis T. Kirsten, H.-H. Do, E. Rahm WG 1, IZBI, University of Leipzig www.izbi.de, dbs.uni-leipzig.de Outline Motivation GeWare Architecture Annotation Integration

Detaljer

Unit Relational Algebra 1 1. Relational Algebra 1. Unit 3.3

Unit Relational Algebra 1 1. Relational Algebra 1. Unit 3.3 Relational Algebra 1 Unit 3.3 Unit 3.3 - Relational Algebra 1 1 Relational Algebra Relational Algebra is : the formal description of how a relational database operates the mathematics which underpin SQL

Detaljer

O v e r o r d n e t m a k r o p e r s p e k t i v p å d i g i t a l e m u l i g h e t e r

O v e r o r d n e t m a k r o p e r s p e k t i v p å d i g i t a l e m u l i g h e t e r O v e r o r d n e t m a k r o p e r s p e k t i v p å d i g i t a l e m u l i g h e t e r S t r a t e g i s k s t y r i n g a v v i r k s o m h e t e r Brit Tone Bergman 19/09/2018 2 En v e r d e n i e

Detaljer

SRP s 4th Nordic Awards Methodology 2018

SRP s 4th Nordic Awards Methodology 2018 SRP s 4th Nordic Awards Methodology 2018 Stockholm 13 September 2018 Awards Methodology 2018 The methodology outlines the criteria by which SRP judges the activity of Manufacturers, Providers and Service

Detaljer

Verifiable Secret-Sharing Schemes

Verifiable Secret-Sharing Schemes Aarhus University Verifiable Secret-Sharing Schemes Irene Giacomelli joint work with Ivan Damgård, Bernardo David and Jesper B. Nielsen Aalborg, 30th June 2014 Verifiable Secret-Sharing Schemes Aalborg,

Detaljer

Emneevaluering GEOV272 V17

Emneevaluering GEOV272 V17 Emneevaluering GEOV272 V17 Studentenes evaluering av kurset Svarprosent: 36 % (5 av 14 studenter) Hvilket semester er du på? Hva er ditt kjønn? Er du...? Er du...? - Annet PhD Candidate Samsvaret mellom

Detaljer

Software applications developed for the maritime service at the Danish Meteorological Institute

Software applications developed for the maritime service at the Danish Meteorological Institute Software applications developed for the maritime service at the Danish Meteorological Institute Anne Marie Munk Jørgensen (ammj@dmi.dk), Ove Kjær, Knud E. Christensen & Morten L. Mortensen Danish Meteorological

Detaljer

1. Explain the language model, what are the weaknesses and strengths of this model?

1. Explain the language model, what are the weaknesses and strengths of this model? Øving 2 Task 1 Language Model 1. Explain the language model, what are the weaknesses and strengths of this model? En language model er en model som brukes til å forenkle spørringer etter ord i dokumenter.

Detaljer

TUNNEL LIGHTING. LED Lighting Technology

TUNNEL LIGHTING. LED Lighting Technology TUNNEL LIGHTING TunLite Linear Designed LED Tunnel & Underpass Light The TunLite is an LED linear luminaire providing reliable solutions to cover the lighting requirements of tunnels and underpasses. It

Detaljer

Trianguleringer og anvendelser

Trianguleringer og anvendelser INF-MAT5370 Trianguleringer og anvendelser Fra seilflysimulatoren Silent Wings Bakgrunn for kurset: Kurset ble til til mens vi vi (foreleserne) arbeidet med oppdrag for industrien på SINTEF. Samtlige deler

Detaljer

Kurskategori 2: Læring og undervisning i et IKT-miljø. vår

Kurskategori 2: Læring og undervisning i et IKT-miljø. vår Kurskategori 2: Læring og undervisning i et IKT-miljø vår Kurs i denne kategorien skal gi pedagogisk og didaktisk kompetanse for å arbeide kritisk og konstruktivt med IKT-baserte, spesielt nettbaserte,

Detaljer

Nærings-PhD i Aker Solutions

Nærings-PhD i Aker Solutions part of Aker Motivasjon og erfaringer Kristin M. Berntsen/Soffi Westin/Maung K. Sein 09.12.2011 2011 Aker Solutions Motivasjon for Aker Solutions Forutsetning Vilje fra bedrift og se nytteverdien av forskning.

Detaljer

Independent Inspection

Independent Inspection Independent Inspection Odd Ivar Johnsen Vidar Nystad Independent Inspection Mål: Felles forståelse og utøvelse av "Independent Inspection" i forbindelse med "Critical Maintenance Task". Independent Inspection

Detaljer

Method validation for NO (10 ppm to 1000 ppm) in nitrogen using the Fischer Rosemount chemiluminescence analyser NOMPUMELELO LESHABANE

Method validation for NO (10 ppm to 1000 ppm) in nitrogen using the Fischer Rosemount chemiluminescence analyser NOMPUMELELO LESHABANE Method validation for NO (10 ppm to 1000 ppm) in nitrogen using the Fischer Rosemount chemiluminescence analyser NOMPUMELELO LESHABANE Presentation Outline Introduction Principle of operation Precision

Detaljer

Capturing the value of new technology How technology Qualification supports innovation

Capturing the value of new technology How technology Qualification supports innovation Capturing the value of new technology How technology Qualification supports innovation Avanserte Marine Operasjoner - Fra operasjon til skip og utstyr Dag McGeorge Ålesund, 1 Contents Introduction - Cheaper,

Detaljer

Læringsmål, vurderingsformer og gradert karakterskala hva er sammenhengen? Om du ønsker, kan du sette inn navn, tittel på foredraget, o.l. her.

Læringsmål, vurderingsformer og gradert karakterskala hva er sammenhengen? Om du ønsker, kan du sette inn navn, tittel på foredraget, o.l. her. 1 Professor Vidar Gynnild: Læringsmål, vurderingsformer og gradert karakterskala hva er sammenhengen? Om du ønsker, kan du sette inn navn, tittel på foredraget, o.l. her. 2 En modell Læringsmål (Intended

Detaljer

HARP-Hybrid Ad Hoc Routing Protocol

HARP-Hybrid Ad Hoc Routing Protocol HARP-Hybrid Ad Hoc Routing Protocol Navid NIKAEIN Christian BONNET Neda NIKAEIN Eurecom Institute Sophia-Antipolis France http://www.eurecom.fr/~nikaeinn 2001 Navid Nikaein Outline ❶ Introduction ❷ Routing

Detaljer

Dyp læring. Sigmund Rolfsjord

Dyp læring. Sigmund Rolfsjord Dyp læring Sigmund Rolfsjord Oversikt 1. Grunnleggende om dyp læring og nevrale nett 2. Konvolusjonsnett 3. Synsfelt med konvolusjonsnett Lær mer: Kurs fra Stanford: http://cs231n.stanford.edu/ Mer inngående

Detaljer

What is is expertise expertise? Individual Individual differ diff ences ences (three (thr ee cent cen r t a r l a lones): easy eas to to test

What is is expertise expertise? Individual Individual differ diff ences ences (three (thr ee cent cen r t a r l a lones): easy eas to to test Expertise in planning & estimation What is it and can one improve it? Jo Hannay (Simula) 1 What is expertise? Individual differences (three central ones): easy to test Personality easy to test Intelligence

Detaljer

Call function of two parameters

Call function of two parameters Call function of two parameters APPLYUSER USER x fµ 1 x 2 eµ x 1 x 2 distinct e 1 0 0 v 1 1 1 e 2 1 1 v 2 2 2 2 e x 1 v 1 x 2 v 2 v APPLY f e 1 e 2 0 v 2 0 µ Evaluating function application The math demands

Detaljer

Endringsdyktige og troverdige systemer

Endringsdyktige og troverdige systemer Endringsdyktige og troverdige systemer Modellering av avhengigheter for å evaluere systemkvalitet 15. jan. 2009 Aida Omerovic SINTEF IKT/UiO 1 Outline Motivation PREDIQT method Practical application of

Detaljer

Comar Benelux NV Brugzavel 8 B-9690 Kluisbergen T +32 (0) F +32 (0)

Comar Benelux NV Brugzavel 8 B-9690 Kluisbergen T +32 (0) F +32 (0) AUTOMATIC CAPACITOR BANKS AAR/5 FOR OUTSTANDING PERFORMANCE AND LONG-TERM VALUE The automatic capacitor banks type AAR/5 offers a unique combination of abilities to give you more convenience, reliability

Detaljer

Databases 1. Extended Relational Algebra

Databases 1. Extended Relational Algebra Databases 1 Extended Relational Algebra Relational Algebra What is an Algebra? Mathematical system consisting of: Operands --- variables or values from which new values can be constructed. Operators ---

Detaljer

FMEM: A Fine- grained Memory Estimator for MapReduce Jobs

FMEM: A Fine- grained Memory Estimator for MapReduce Jobs FMEM: A Fine- grained Memory Estimator for MapReduce Jobs Lijie Xu 1,2, Jie Liu 1, and Jun Wei 1 1 Institute of Software, Chinese Academy of Sciences 2 University of Chinese Academy of Sciences 6/26/2013

Detaljer

Generalization of age-structured models in theory and practice

Generalization of age-structured models in theory and practice Generalization of age-structured models in theory and practice Stein Ivar Steinshamn, stein.steinshamn@snf.no 25.10.11 www.snf.no Outline How age-structured models can be generalized. What this generalization

Detaljer

Ringvorlesung Biophysik 2016

Ringvorlesung Biophysik 2016 Ringvorlesung Biophysik 2016 Born-Oppenheimer Approximation & Beyond Irene Burghardt (burghardt@chemie.uni-frankfurt.de) http://www.theochem.uni-frankfurt.de/teaching/ 1 Starting point: the molecular Hamiltonian

Detaljer

SIU Retningslinjer for VET mobilitet

SIU Retningslinjer for VET mobilitet SIU Retningslinjer for VET mobilitet Gardermoen, 16.09.2014 Oppstart- og erfaringsseminar Tore Kjærgård Carl Endre Espeland 2 Kort om Erasmus+ EUs utdanningsprogram for perioden 2014 2020 Budsjett: 14,7

Detaljer

Exploratory Analysis of a Large Collection of Time-Series Using Automatic Smoothing Techniques

Exploratory Analysis of a Large Collection of Time-Series Using Automatic Smoothing Techniques Exploratory Analysis of a Large Collection of Time-Series Using Automatic Smoothing Techniques Ravi Varadhan, Ganesh Subramaniam Johns Hopkins University AT&T Labs - Research 1 / 28 Introduction Goal:

Detaljer

NO X -chemistry modeling for coal/biomass CFD

NO X -chemistry modeling for coal/biomass CFD NO X -chemistry modeling for coal/biomass CFD Jesper Møller Pedersen 1, Larry Baxter 2, Søren Knudsen Kær 3, Peter Glarborg 4, Søren Lovmand Hvid 1 1 DONG Energy, Denmark 2 BYU, USA 3 AAU, Denmark 4 DTU,

Detaljer

Confidence-based Data Management for Personal Area Sensor Nets

Confidence-based Data Management for Personal Area Sensor Nets Confidence-based Data Management for Personal Area Sensor Nets Nesime Tatbul, Stan Zdonik Brown University Mark Buller, Reed Hoyt, Steve Mullen USARIEM Talk Outline Warfighter Physiologic Status Monitoring

Detaljer

Sikkert Drillingnettverk på CAT-D Rig

Sikkert Drillingnettverk på CAT-D Rig Sikkert Drillingnettverk på CAT-D Rig Med fokus på IT sikkerhet i offshore bransjen Kristiansand, 21/10/2014, Asgeir Skretting, Dag Tang Sikkert Drillingnettverk på CAT-D Rig Agenda Hvorfor sikker offshore

Detaljer

Hva slags AAR-krav i framtida? Begrunnelse for Felt/Lab Performance. COIN fagdag 20. mai 2008 Terje F. Rønning, Norcem AS

Hva slags AAR-krav i framtida? Begrunnelse for Felt/Lab Performance. COIN fagdag 20. mai 2008 Terje F. Rønning, Norcem AS Hva slags AAR-krav i framtida? Begrunnelse for Felt/Lab Performance COIN fagdag 20. mai 2008 Terje F. Rønning, Norcem AS Rilem forslag til CEN Level of Precaution P2: A normal level of precaution against

Detaljer

6 December 2011 DG CLIMA. Stakeholder meeting on LDV CO 2 emissions - Scene setter

6 December 2011 DG CLIMA. Stakeholder meeting on LDV CO 2 emissions - Scene setter 6 December 2011 DG CLIMA 1 Stakeholder meeting on LDV CO 2 emissions - Scene setter Context of 80-95% reduction 2 Keeping average global temperature increase below 2 C confirmed as global objective (UNFCCC-

Detaljer

Understanding Social and Environmental Conflicts in Mining Exploration Simexmin May 18, 2016 Alan Dabbs Social Capital Group. Tia Maria Conflict

Understanding Social and Environmental Conflicts in Mining Exploration Simexmin May 18, 2016 Alan Dabbs Social Capital Group. Tia Maria Conflict Understanding Social and Environmental Conflicts in Mining Exploration Simexmin May 18, 2016 Alan Dabbs Social Capital Group Tia Maria Conflict Agenda Perspective Drivers of social conflict Points of Interaction

Detaljer

On Capacity Planning for Minimum Vulnerability

On Capacity Planning for Minimum Vulnerability On Capacity Planning for Minimum Vulnerability Alireza Bigdeli Ali Tizghadam Alberto Leon-Garcia University of Toronto DRCN - October 2011 Kakow - Poland 1 Outline Introduction Network Criticality and

Detaljer

SAS FANS NYTT & NYTTIG FRA VERKTØYKASSA TIL SAS 4. MARS 2014, MIKKEL SØRHEIM

SAS FANS NYTT & NYTTIG FRA VERKTØYKASSA TIL SAS 4. MARS 2014, MIKKEL SØRHEIM SAS FANS NYTT & NYTTIG FRA VERKTØYKASSA TIL SAS 4. MARS 2014, MIKKEL SØRHEIM 2 TEMA 1 MULTIPROSESSERING MED DATASTEGET Multiprosessering har lenge vært et tema i SAS Stadig ny funksjonalitet er med på

Detaljer

Quality Policy. HSE Policy

Quality Policy. HSE Policy 1 2 Quality Policy HSE Policy Astra North shall provide its customers highly motivated personnel with correct competence and good personal qualities to each specific assignment. Astra North believes a

Detaljer

The building blocks of a biogas strategy

The building blocks of a biogas strategy The building blocks of a biogas strategy Presentation of the report «Background report for a biogas strategy» («Underlagsmateriale til tverrsektoriell biogass-strategi») Christine Maass, Norwegian Environment

Detaljer

KROPPEN LEDER STRØM. Sett en finger på hvert av kontaktpunktene på modellen. Da får du et lydsignal.

KROPPEN LEDER STRØM. Sett en finger på hvert av kontaktpunktene på modellen. Da får du et lydsignal. KROPPEN LEDER STRØM Sett en finger på hvert av kontaktpunktene på modellen. Da får du et lydsignal. Hva forteller dette signalet? Gå flere sammen. Ta hverandre i hendene, og la de to ytterste personene

Detaljer

EFFEKTIV BRUK AV VIDEO I TRENING OG FORSKNING

EFFEKTIV BRUK AV VIDEO I TRENING OG FORSKNING Canon XL-1s med en Canon EF 100-400 mm objektiv Robert C. Reid, PhD Norges skiforbund hastighet bruk av video Før Filming: Hva er du ut etter å belyse? Hvordan vil du bruke opptaket etterpå (bilde sekvens,

Detaljer

How Bridges Work Sgrad 2001

How Bridges Work Sgrad 2001 How Bridges Work Sgrad 2001 The Basic s There are three major types of bridges: The beam bridge The arch bridge The suspension bridge prepared by Mr.S.Grad 2 The biggest difference between the three is

Detaljer

Harmonisation of terminological resources future perspectives

Harmonisation of terminological resources future perspectives Harmonisation of terminological resources future perspectives CLARA Terminology course Gisle Andersen, gisle.andersen@nhh.no 17. September 2010 www.nhh.no Utfordringer og målsettinger Challenges å utvikle

Detaljer

STILLAS - STANDARD FORSLAG FRA SEF TIL NY STILLAS - STANDARD

STILLAS - STANDARD FORSLAG FRA SEF TIL NY STILLAS - STANDARD FORSLAG FRA SEF TIL NY STILLAS - STANDARD 1 Bakgrunnen for dette initiativet fra SEF, er ønsket om å gjøre arbeid i høyden tryggere / sikrere. Både for stillasmontører og brukere av stillaser. 2 Reviderte

Detaljer

Oppgave 1a Definer følgende begreper: Nøkkel, supernøkkel og funksjonell avhengighet.

Oppgave 1a Definer følgende begreper: Nøkkel, supernøkkel og funksjonell avhengighet. TDT445 Øving 4 Oppgave a Definer følgende begreper: Nøkkel, supernøkkel og funksjonell avhengighet. Nøkkel: Supernøkkel: Funksjonell avhengighet: Data i en database som kan unikt identifisere (et sett

Detaljer

ENERGY STAR QUALIFIED UPS PRODUCTS

ENERGY STAR QUALIFIED UPS PRODUCTS ENERGY STAR QUALIFIED UPS PRODUCTS MISSION The ENERGY STAR Program was established by the U.S. Environmental Protection Agency (EPA) as a way to identify and promote energy-efficient products. Its goal

Detaljer

SERK1/2 Acts as a Partner of EMS1 to Control Anther Cell Fate Determination in Arabidopsis

SERK1/2 Acts as a Partner of EMS1 to Control Anther Cell Fate Determination in Arabidopsis SERK1/2 Acts as a Partner of EMS1 to Control Anther Cell Fate Determination in Arabidopsis Supplemental Data Supplemental Figure S1 Supplemental Figure S1. Identification of the ems1-2 weak allele. A,

Detaljer

Innovasjonsvennlig anskaffelse

Innovasjonsvennlig anskaffelse UNIVERSITETET I BERGEN Universitetet i Bergen Innovasjonsvennlig anskaffelse Fredrikstad, 20 april 2016 Kjetil Skog 1 Universitetet i Bergen 2 Universitetet i Bergen Driftsinntekter på 4 milliarder kr

Detaljer

Slope-Intercept Formula

Slope-Intercept Formula LESSON 7 Slope Intercept Formula LESSON 7 Slope-Intercept Formula Here are two new words that describe lines slope and intercept. The slope is given by m (a mountain has slope and starts with m), and intercept

Detaljer

Improving Coarsening Schemes for Hypergraph Partitioning by Exploiting Community Structure

Improving Coarsening Schemes for Hypergraph Partitioning by Exploiting Community Structure Improving Coarsening Schemes for Hypergraph Partitioning by Exploiting Community Structure SEA 17 June 23, 217 Tobias Heuer and Sebastian Schlag I NSTITUTE OF T HEORETICAL I NFORMATICS A LGORITHMICS G

Detaljer

Risikofokus - også på de områdene du er ekspert

Risikofokus - også på de områdene du er ekspert Risikofokus - også på de områdene du er ekspert - hvordan kan dette se ut i praksis? - Ingen er for gammel til å begå nye dumheter Nytt i ISO 9001:2015 Vokabular Kontekst Dokumentasjonskrav Lederskap Stategi-politikk-mål

Detaljer

KEEPING THE WORLD MOVING. Ferske data til navigasjonssystemer fra datafangst til sluttbruker

KEEPING THE WORLD MOVING. Ferske data til navigasjonssystemer fra datafangst til sluttbruker KEEPING THE WORLD MOVING Ferske data til navigasjonssystemer fra datafangst til sluttbruker Our vision Our vision is a safe, connected, autonomous world, free of congestion and emissions. Our mission We

Detaljer

Dagens tema: Eksempel Klisjéer (mønstre) Tommelfingerregler

Dagens tema: Eksempel Klisjéer (mønstre) Tommelfingerregler UNIVERSITETET I OSLO INF1300 Introduksjon til databaser Dagens tema: Eksempel Klisjéer (mønstre) Tommelfingerregler Institutt for informatikk Dumitru Roman 1 Eksempel (1) 1. The system shall give an overview

Detaljer

Stationary Phase Monte Carlo Methods

Stationary Phase Monte Carlo Methods Stationary Phase Monte Carlo Methods Daniel Doro Ferrante G. S. Guralnik, J. D. Doll and D. Sabo HET Physics Dept, Brown University, USA. danieldf@het.brown.edu www.het.brown.edu Introduction: Motivations

Detaljer

Molare forsterkningsbetingelser

Molare forsterkningsbetingelser Molare forsterkningsbetingelser Hva er mekanismen(e) bak forsterkning? Hvor langt opp eller ned skal man skru mikroskopet for å se godt nok? Kjetil Viken 1 2 ARBEIDSDAG sitte ved pc formelle samtaler møter

Detaljer

Energy Calibration for the Forward Detector at WASA-at-COSY

Energy Calibration for the Forward Detector at WASA-at-COSY Energy Calibration for the Forward Detector at WASA-at-COSY Kay Demmich Westfälische Wilhelms-Universität Münster, Institut für Kernphysik DPG Spring Meeting (HK 42.7) 5. März 23 K. Demmich (WWU) Calibration

Detaljer

BioCarb+ NFR KPN prosjekt MNOK. Enabling the biocarbon value chain for energy

BioCarb+ NFR KPN prosjekt MNOK. Enabling the biocarbon value chain for energy Enabling the biocarbon value chain for energy BioCarb+ Dr. Ing. Øyvind Skreiberg Sjefforsker, SINTEF Energi AS BioCarb+ prosjektleder oyvind.skreiberg@sintef.no http://www.sintef.no/biocarb NFR KPN prosjekt

Detaljer

Level Set methods. Sandra Allaart-Bruin. Level Set methods p.1/24

Level Set methods. Sandra Allaart-Bruin. Level Set methods p.1/24 Level Set methods Sandra Allaart-Bruin sbruin@win.tue.nl Level Set methods p.1/24 Overview Introduction Level Set methods p.2/24 Overview Introduction Boundary Value Formulation Level Set methods p.2/24

Detaljer

Jacob Nielsens 10 heuristikker. Dansk Guru med 10 Tommelfingerregler en_(usability_consultant)

Jacob Nielsens 10 heuristikker. Dansk Guru med 10 Tommelfingerregler  en_(usability_consultant) Jacob Nielsens 10 heuristikker Dansk Guru med 10 Tommelfingerregler http://en.wikipedia.org/wiki/jakob_niels en_(usability_consultant) Jacob Nilsens heuristikker - 1 Visibility of system status The system

Detaljer

Gaute Langeland September 2016

Gaute Langeland September 2016 Gaute Langeland September 2016 Svak krone 10,4 10 9,6 9,2 8,8 8,4 EURNOK 8 7,6 7,2 6,8 3jan00 3jan02 3jan04 3jan06 3jan08 3jan10 3jan12 3jan14 3jan16 2 12.10.2016 Ikke helt tilfeldig 3 12.10.2016 Hvordan

Detaljer

Jeroen Stil Institute for Space Imaging Science. University of Calgary

Jeroen Stil Institute for Space Imaging Science. University of Calgary Jeroen Stil Institute for Space Imaging Science University of Calgary Origin and evolution of cosmic magnetic fields is a key science goal for the SKA Rotation measures of polarized background sources

Detaljer

REMOVE CONTENTS FROM BOX. VERIFY ALL PARTS ARE PRESENT READ INSTRUCTIONS CAREFULLY BEFORE STARTING INSTALLATION

REMOVE CONTENTS FROM BOX. VERIFY ALL PARTS ARE PRESENT READ INSTRUCTIONS CAREFULLY BEFORE STARTING INSTALLATION 2011-2014 FORD EXPLORER PARTS LIST Qty Part Description Qty Part Description 1 Bull Bar 2 12mm x 35mm Bolt Plates 1 Passenger/Right Mounting Bracket 2 12mm Nut Plate 1 Driver/Left Mounting Bracket 2 12mm

Detaljer

Har vi forretningsmodeller som muliggjør effektiv utvikling og introduksjon av nye tjenester i helsesektoren?

Har vi forretningsmodeller som muliggjør effektiv utvikling og introduksjon av nye tjenester i helsesektoren? Odd Arild Lehne, Advisor Innovation Projects, Oslo Medtech Har vi forretningsmodeller som muliggjør effektiv utvikling og introduksjon av nye tjenester i helsesektoren? Oslo Medtech facts & figures Founded

Detaljer

ENERGY STAR Qualified UPS Products

ENERGY STAR Qualified UPS Products The ENERGY STAR Program was established by the U.S. Environmental Protection Agency (EPA) as a way to identify and promote energy-efficient products, in order to reduce energy consumption, help individuals

Detaljer

INTELLIGENT TEKNOLOGISK LIV

INTELLIGENT TEKNOLOGISK LIV INTELLIGENT TEKNOLOGISK LIV Mennesket som utgangspunkt AVGRENSNING AVGRENSNING Teknologi: "The application of scientific knowledge for practical purposes, especially in industry" (oxford dictionary). AVGRENSNING

Detaljer

Explicit vs. Implicit Polymorphism in OML. Thomas Christensen MSc. (CS) (Soon) University of Aarhus, Denmark

Explicit vs. Implicit Polymorphism in OML. Thomas Christensen MSc. (CS) (Soon) University of Aarhus, Denmark Explicit vs. Implicit Polymorphism in OML Thomas Christensen MSc. (CS) (Soon) University of Aarhus, Denmark Agenda Explicit vs. Implicit polymorphism in OML Type inference Problems Generics in OML Conclusion

Detaljer

EXAM TTM4128 SERVICE AND RESOURCE MANAGEMENT EKSAM I TTM4128 TJENESTE- OG RESSURSADMINISTRASJON

EXAM TTM4128 SERVICE AND RESOURCE MANAGEMENT EKSAM I TTM4128 TJENESTE- OG RESSURSADMINISTRASJON Side 1 av 5 NTNU Norges teknisk-naturvitenskapelige universitet Institutt for telematikk EXAM TTM4128 SERVICE AND RESOURCE MANAGEMENT EKSAM I TTM4128 TJENESTE- OG RESSURSADMINISTRASJON Contact person /

Detaljer

Q2 Results July 17, Hans Stråberg President and CEO. Fredrik Rystedt CFO

Q2 Results July 17, Hans Stråberg President and CEO. Fredrik Rystedt CFO Q2 Results 2007 July 17, 2007 Hans Stråberg President and CEO Fredrik Rystedt CFO Q2 Highlights EBIT (SEKb) EBIT margin (%) 2.5 2 1.5 1 0.5 0 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 8% 7% 6% 5% 4% 3% 2% 1% 0% Group

Detaljer

Tangible Interaction. Tone Bratteteig

Tangible Interaction. Tone Bratteteig Tangible Interaction Tone Bratteteig in1060: 23/4 2018 Hornecker & Buur CHI 2006 Proceedings Designing for Tangible Interactions April 22-27, 2006 Montréal, Québec, Canada Getting a Grip on Tangible Interaction:

Detaljer

Hvordan jobber reiselivsgründere med sine etableringer? Sølvi Solvoll Klyngesamling, Bodø

Hvordan jobber reiselivsgründere med sine etableringer? Sølvi Solvoll Klyngesamling, Bodø Hvordan jobber reiselivsgründere med sine etableringer? Sølvi Solvoll Klyngesamling, Bodø 14.02.2018 Hvilke beslutninger har du tatt i dag? Planlegge eller effektuere? Effectuation; måten ekspertgründeren

Detaljer

// Translation // KLART SVAR «Free-Range Employees»

// Translation // KLART SVAR «Free-Range Employees» // Translation // KLART SVAR «Free-Range Employees» Klart Svar is a nationwide multiple telecom store, known as a supplier of mobile phones and wireless office solutions. The challenge was to make use

Detaljer

Interventions in the Cerebral palsy follow-up program: Reidun Jahnsen, PT PhD

Interventions in the Cerebral palsy follow-up program: Reidun Jahnsen, PT PhD Interventions in the Cerebral palsy follow-up program: Reidun Jahnsen, PT PhD Subgroups of CP CPOP 214 n=188 Gross Motor Function Classification System Palisano 1997, CPOP 214 171 CPOP 213 n=856 Manual

Detaljer

E-Learning Design. Speaker Duy Hai Nguyen, HUE Online Lecture

E-Learning Design. Speaker Duy Hai Nguyen, HUE Online Lecture E-Learning Design Speaker Duy Hai Nguyen, HUE Online Lecture Design Educational Design Navigation Design Educational Design Some Important Considerations: 1. Authentic learning environment: For effective

Detaljer

WÄRTSILÄ MARINE SOLUTION POWER CONVERSION INNOVATIVE LAV- OG NULLUTSLIPPSLØSNINGER OG UTFORDRINGER MED Å FÅ DISSE INN I MARKEDET.

WÄRTSILÄ MARINE SOLUTION POWER CONVERSION INNOVATIVE LAV- OG NULLUTSLIPPSLØSNINGER OG UTFORDRINGER MED Å FÅ DISSE INN I MARKEDET. INNOVATIVE LAV- OG NULLUTSLIPPSLØSNINGER OG UTFORDRINGER MED Å FÅ DISSE INN I MARKEDET. WÄRTSILÄ MARINE SOLUTION POWER CONVERSION INGVE SØRFONN 1 THE FUTURE IS NOW! 2 FROM PRODUCT TO ECOSYSTEM 3 READY

Detaljer