SlideShare une entreprise Scribd logo
1  sur  34
Technology Overview
Chris Slaughter
President
Prof. Sriram Vishwanath
Chief Scientist
Our Team                                                   2




           Prof. Sriram Vishwanath
           • 9 years, Prof., UT Austin
           • Information Theory, Entrepreneurship
           • 2 Robotics Labs: MA coordination and 3D
              Perception

           Chris Slaughter
           • PhD. Candidate, Electrical Engineering
           • 1.5 years consulting T.O. of Austin Startup
           • Research Lead, UT Perception Laboratory

           Ongoing Collaborations/Partnerships
           • Lockheed Martin
           • HKS
           • NVIDIA
           • UC Berkeley
Our Team                                                                           3




•   Computer Vision: Multi-view geometry and stereo; tracking; 3d reconstruction

•   High Performance Computing: General purpose graphics programming
    (GPGPU), parallel/distributed computing, heterogeneous computation

•   Statistics and Learning: Large scale clustering problems, compressive motion
    analysis, graphical inference

•   Embedded Systems: Board design, multi-processor
    interaction, interfaces, power/weight/form factor
Mission Statement                                             4




        To Teach Unmanned Vehicles to See as Humans Do


                              Applications:

  •   Absolute localization in GPS denied scenarios
  •   Visual tracking odometry
  •   Landmark detection and landmark-based navigation
  •   Terrain mapping and change detection for IED disposal
  •   Immersive visualization for situational awareness
  •   Disaster response
System Architecture                                   4




                      Server Nodes




     Producer Nodes                  Consumer Nodes
System Architecture                                   5




                      Server Nodes




     Producer Nodes                  Consumer Nodes
Producer Nodes
Producer Nodes                                                    7




               Computing Element:
            Embedded GPGPU processor
                 ARM multi-core
            Heterogeneous architecture



                                         Optoelectronic Device:
                                           Active Stereo (IR)
                                             Passive Stereo
                                              Laser / TOF
Producer Nodes                                                                    8




                                         Key Features:
                                         • Visual Odometry ( < 3 mm )
                                         • Mapping
                                         • Landmark Extraction
                                         • High Data Rates ( 9.2M pts / sec )
                                         • Volumetric / Manifold Reconstruction
                                         • Source Compression for Uplink



 Compatibility:
 • Low power ( < 4.5 W )
 • Low weight ( cell phone – battery )
 • Low cost ( COTS sensors )
Producer Nodes                          9




          Signal Processing
  • Fundamental task in vision-based
    algorithms
  • Most algorithms too slow even for
    desktops

  •   Bilateral filters
  •   Pyramid computation
  •   Depth-RGB conversion
  •   SIFT descriptors
  •   Object recognition
Image Filtering
Pyramid Computation
Frame Matching
Producer Nodes                        1
                                      3




         Visual Odometry
  • RGB-based
     • 30 FPS
     • Accuracy in mm
     • Compressive sensing solution

  • Range-based
     • 9.2M pts / sec
     • 90T comps / sec
     • Towards drift free?
Vision-Only Tracking
RANSAC (1983)
RANSAC (1983)




 SOLO (2010)
Producer Nodes                                                         1
                                                                       7




  Can a mobile device map 3D environments to produce dense
        3D data points rather than sparse landmarks?


                           Dense Reconstruction
  • Landmark-based mapping (1960’s – present) – SOLVED
      • DARPA grand challenge
  • Efficient global alignment (1999 – 2005) – SOLVED
  • Multiple view geometry + dense reconstruction (1980’s – present)
  • Live dense reconstruction (2006 – present)
      • KinectFusion
      • Range-based SLAM
      • Dense Tracking and Mapping
Volumetric Reconstruction
Point Cloud Reconstruction
Patchwork Reconstruction
Depth Data
Producer Nodes                             2
                                           2




      Patchwork Reconstructions
  •   More memory efficient than volumes
  •   Faster integration and tracking
  •   Efficient caching interplay
  •   Global refinement
  •   Runs on embedded device
Server Nodes
Server Nodes                             2
                                         4




         Global Refinement
  • Maps must be globally consistent
  • Dense reconstruction doesn’t allow
    for this refinement
  • Patchwork generalizes to this
    functionality
  • Inherently multi-core problem
  • ARM architecture for GraDeS
No Refinement




HOGMAN (2007)
Full Visualization                         2
                                           6




        Custom Visualization
  •   Key task: visualization
  •   Interactivity (XBOX controller)
  •   Full support for 3d / 2d streams
  •   Event processing
  •   Interoperability with GPGPU (CUDA)
  •   Compression for consumer nodes
  •   Fully scalable


          Rendering Pipeline
  • Custom rendering pipeline for
    Patchwork
  • Volumetric and Point Cloud
  • Raycasting with lighting sources
Consumer Nodes
Consumer Nodes                                                                       2
                                                                                     8




    Our technology can produce maps at high speeds and unprecedented fidelities
                    But.. What   to do with this content?

        Visual Localization                      Situational Awareness
• Localization a major problem in GPS-     • Visualize mapping assets in real time
  denied scenarios                           from cell phones
    • “Urban canyons”                      • Coordinate with server and receive
    • Indoor environments                    compressed video stream
    • MAV / UGV coordination               • Back-end models dynamics of
• Existing solutions based mostly on         adversaries
  state estimation                         • Extensible visualizer: new
• Possible to query large maps for           tags, models, data sources
  location?
Visual Absolute Localization
Feasible Path   Infeasible Path
Positional Decoding
Conclusion                                                                 3
                                                                           2




             Current trends in computer vision and robotics:
             • High performance computing
             • Live dense reconstruction
             • Range-based tracking and mapping

             Our architecture:
             • Producer nodes:
                 • COTS sensors
                 • Commodity computational unit
                 • Dense tracking and mapping
             • Server nodes
                 • Combine producer data into large maps
                 • Serve consumer nodes
             • Consumer nodes
                 • Visual absolute localization and remote visualization
Thanks!

Contenu connexe

Tendances

"Digital Gimbal: Rock-steady Video Stabilization without Extra Weight!," a Pr...
"Digital Gimbal: Rock-steady Video Stabilization without Extra Weight!," a Pr..."Digital Gimbal: Rock-steady Video Stabilization without Extra Weight!," a Pr...
"Digital Gimbal: Rock-steady Video Stabilization without Extra Weight!," a Pr...
Edge AI and Vision Alliance
 
A Vision-Based Mobile Platform for Seamless Indoor/Outdoor Positioning
A Vision-Based Mobile Platform for Seamless Indoor/Outdoor PositioningA Vision-Based Mobile Platform for Seamless Indoor/Outdoor Positioning
A Vision-Based Mobile Platform for Seamless Indoor/Outdoor Positioning
Guillaume Gales
 

Tendances (8)

GOAR: GIS Oriented Mobile Augmented Reality for Urban Landscape Assessment
GOAR: GIS Oriented Mobile Augmented Reality for Urban Landscape AssessmentGOAR: GIS Oriented Mobile Augmented Reality for Urban Landscape Assessment
GOAR: GIS Oriented Mobile Augmented Reality for Urban Landscape Assessment
 
"Digital Gimbal: Rock-steady Video Stabilization without Extra Weight!," a Pr...
"Digital Gimbal: Rock-steady Video Stabilization without Extra Weight!," a Pr..."Digital Gimbal: Rock-steady Video Stabilization without Extra Weight!," a Pr...
"Digital Gimbal: Rock-steady Video Stabilization without Extra Weight!," a Pr...
 
SOAR: SENSOR ORIENTED MOBILE AUGMENTED REALITY FOR URBAN LANDSCAPE ASSESSMENT
SOAR: SENSOR ORIENTED MOBILE AUGMENTED REALITY FOR URBAN LANDSCAPE ASSESSMENTSOAR: SENSOR ORIENTED MOBILE AUGMENTED REALITY FOR URBAN LANDSCAPE ASSESSMENT
SOAR: SENSOR ORIENTED MOBILE AUGMENTED REALITY FOR URBAN LANDSCAPE ASSESSMENT
 
DISTRIBUTED AND SYNCHRONISED VR MEETING USING CLOUD COMPUTING: Availability a...
DISTRIBUTED AND SYNCHRONISED VR MEETING USING CLOUD COMPUTING: Availability a...DISTRIBUTED AND SYNCHRONISED VR MEETING USING CLOUD COMPUTING: Availability a...
DISTRIBUTED AND SYNCHRONISED VR MEETING USING CLOUD COMPUTING: Availability a...
 
Lift using projected coded light for finger tracking and device augmentation
Lift using projected coded light for finger tracking and device augmentationLift using projected coded light for finger tracking and device augmentation
Lift using projected coded light for finger tracking and device augmentation
 
An Authoring Solution for a Façade-Based AR Platform: Infrastructure, Annota...
An Authoring Solution for  a Façade-Based AR Platform: Infrastructure, Annota...An Authoring Solution for  a Façade-Based AR Platform: Infrastructure, Annota...
An Authoring Solution for a Façade-Based AR Platform: Infrastructure, Annota...
 
A Vision-Based Mobile Platform for Seamless Indoor/Outdoor Positioning
A Vision-Based Mobile Platform for Seamless Indoor/Outdoor PositioningA Vision-Based Mobile Platform for Seamless Indoor/Outdoor Positioning
A Vision-Based Mobile Platform for Seamless Indoor/Outdoor Positioning
 
A maskless exposure device for rapid photolithographic prototyping of sensor ...
A maskless exposure device for rapid photolithographic prototyping of sensor ...A maskless exposure device for rapid photolithographic prototyping of sensor ...
A maskless exposure device for rapid photolithographic prototyping of sensor ...
 

En vedette (7)

Complex Terrain and Snow Modeling: Vehicle - Terrain Interaction
Complex Terrain and Snow Modeling: Vehicle - Terrain InteractionComplex Terrain and Snow Modeling: Vehicle - Terrain Interaction
Complex Terrain and Snow Modeling: Vehicle - Terrain Interaction
 
Future Research Directions in Military Ground Vehicle Mobility
Future Research Directions in Military Ground Vehicle MobilityFuture Research Directions in Military Ground Vehicle Mobility
Future Research Directions in Military Ground Vehicle Mobility
 
IEDs awareness
IEDs awarenessIEDs awareness
IEDs awareness
 
Ied vehicle search
Ied vehicle searchIed vehicle search
Ied vehicle search
 
New microsoft power point presentation (2)
New microsoft power point presentation (2)New microsoft power point presentation (2)
New microsoft power point presentation (2)
 
tardec report
tardec reporttardec report
tardec report
 
Blast technologies wiaman
Blast technologies wiamanBlast technologies wiaman
Blast technologies wiaman
 

Similaire à TARDEC Presentation 2

“AI-ISP: Adding Real-time AI Functionality to Image Signal Processing with Re...
“AI-ISP: Adding Real-time AI Functionality to Image Signal Processing with Re...“AI-ISP: Adding Real-time AI Functionality to Image Signal Processing with Re...
“AI-ISP: Adding Real-time AI Functionality to Image Signal Processing with Re...
Edge AI and Vision Alliance
 
Compact Descriptors for Visual Search
Compact Descriptors for Visual SearchCompact Descriptors for Visual Search
Compact Descriptors for Visual Search
Antonio Capone
 
“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...
“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...
“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...
Edge AI and Vision Alliance
 
Bringing Wireless Sensing to its full potential
Bringing Wireless Sensing to its full potentialBringing Wireless Sensing to its full potential
Bringing Wireless Sensing to its full potential
Adrian Hornsby
 
"The Coming Shift from Image Sensors to Image Sensing," a Presentation from LG
"The Coming Shift from Image Sensors to Image Sensing," a Presentation from LG"The Coming Shift from Image Sensors to Image Sensing," a Presentation from LG
"The Coming Shift from Image Sensors to Image Sensing," a Presentation from LG
Edge AI and Vision Alliance
 

Similaire à TARDEC Presentation 2 (20)

Elevation mapping using stereo vision enabled heterogeneous multi-agent robot...
Elevation mapping using stereo vision enabled heterogeneous multi-agent robot...Elevation mapping using stereo vision enabled heterogeneous multi-agent robot...
Elevation mapping using stereo vision enabled heterogeneous multi-agent robot...
 
Software Architecture For Condition Monitoring Of Mobile Underground
Software Architecture For Condition Monitoring Of Mobile UndergroundSoftware Architecture For Condition Monitoring Of Mobile Underground
Software Architecture For Condition Monitoring Of Mobile Underground
 
Emerging vision technologies
Emerging vision technologiesEmerging vision technologies
Emerging vision technologies
 
Gpu with cuda architecture
Gpu with cuda architectureGpu with cuda architecture
Gpu with cuda architecture
 
“AI-ISP: Adding Real-time AI Functionality to Image Signal Processing with Re...
“AI-ISP: Adding Real-time AI Functionality to Image Signal Processing with Re...“AI-ISP: Adding Real-time AI Functionality to Image Signal Processing with Re...
“AI-ISP: Adding Real-time AI Functionality to Image Signal Processing with Re...
 
Hai Tao at AI Frontiers: Deep Learning For Embedded Vision System
Hai Tao at AI Frontiers: Deep Learning For Embedded Vision SystemHai Tao at AI Frontiers: Deep Learning For Embedded Vision System
Hai Tao at AI Frontiers: Deep Learning For Embedded Vision System
 
Compact Descriptors for Visual Search
Compact Descriptors for Visual SearchCompact Descriptors for Visual Search
Compact Descriptors for Visual Search
 
Real-time DeepLearning on IoT Sensor Data
Real-time DeepLearning on IoT Sensor DataReal-time DeepLearning on IoT Sensor Data
Real-time DeepLearning on IoT Sensor Data
 
MIPI DevCon 2016: MIPI CSI-2 Application for Vision and Sensor Fusion Systems
MIPI DevCon 2016: MIPI CSI-2 Application for Vision and Sensor Fusion SystemsMIPI DevCon 2016: MIPI CSI-2 Application for Vision and Sensor Fusion Systems
MIPI DevCon 2016: MIPI CSI-2 Application for Vision and Sensor Fusion Systems
 
“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...
“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...
“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...
 
Performance Evaluation and Comparison of Service-based Image Processing based...
Performance Evaluation and Comparison of Service-based Image Processing based...Performance Evaluation and Comparison of Service-based Image Processing based...
Performance Evaluation and Comparison of Service-based Image Processing based...
 
Understanding the world in 3D with AI.pdf
Understanding the world in 3D with AI.pdfUnderstanding the world in 3D with AI.pdf
Understanding the world in 3D with AI.pdf
 
System-on-Chip Programmable Retina
System-on-Chip Programmable RetinaSystem-on-Chip Programmable Retina
System-on-Chip Programmable Retina
 
Applying Deep Learning Vision Technology to low-cost/power Embedded Systems
Applying Deep Learning Vision Technology to low-cost/power Embedded SystemsApplying Deep Learning Vision Technology to low-cost/power Embedded Systems
Applying Deep Learning Vision Technology to low-cost/power Embedded Systems
 
Efficient architecture to condensate visual information driven by attention ...
Efficient architecture to condensate visual information driven by attention ...Efficient architecture to condensate visual information driven by attention ...
Efficient architecture to condensate visual information driven by attention ...
 
Bringing Wireless Sensing to its full potential
Bringing Wireless Sensing to its full potentialBringing Wireless Sensing to its full potential
Bringing Wireless Sensing to its full potential
 
Spark Technology Center IBM
Spark Technology Center IBMSpark Technology Center IBM
Spark Technology Center IBM
 
"The Coming Shift from Image Sensors to Image Sensing," a Presentation from LG
"The Coming Shift from Image Sensors to Image Sensing," a Presentation from LG"The Coming Shift from Image Sensors to Image Sensing," a Presentation from LG
"The Coming Shift from Image Sensors to Image Sensing," a Presentation from LG
 
Operationalizing Machine Learning Using GPU-accelerated, In-database Analytics
Operationalizing Machine Learning Using GPU-accelerated, In-database AnalyticsOperationalizing Machine Learning Using GPU-accelerated, In-database Analytics
Operationalizing Machine Learning Using GPU-accelerated, In-database Analytics
 
OpenStreetMap in 3D - current developments
OpenStreetMap in 3D - current developmentsOpenStreetMap in 3D - current developments
OpenStreetMap in 3D - current developments
 

Dernier

Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and Myths
Joaquim Jorge
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI Solutions
Enterprise Knowledge
 
Histor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slideHistor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slide
vu2urc
 

Dernier (20)

Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processors
 
Presentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreterPresentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreter
 
Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024
 
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfThe Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024
 
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationFrom Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
 
Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and Myths
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonets
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
 
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
 
Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed texts
 
What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI Solutions
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
 
Histor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slideHistor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slide
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organization
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected Worker
 
presentation ICT roal in 21st century education
presentation ICT roal in 21st century educationpresentation ICT roal in 21st century education
presentation ICT roal in 21st century education
 

TARDEC Presentation 2

  • 1. Technology Overview Chris Slaughter President Prof. Sriram Vishwanath Chief Scientist
  • 2. Our Team 2 Prof. Sriram Vishwanath • 9 years, Prof., UT Austin • Information Theory, Entrepreneurship • 2 Robotics Labs: MA coordination and 3D Perception Chris Slaughter • PhD. Candidate, Electrical Engineering • 1.5 years consulting T.O. of Austin Startup • Research Lead, UT Perception Laboratory Ongoing Collaborations/Partnerships • Lockheed Martin • HKS • NVIDIA • UC Berkeley
  • 3. Our Team 3 • Computer Vision: Multi-view geometry and stereo; tracking; 3d reconstruction • High Performance Computing: General purpose graphics programming (GPGPU), parallel/distributed computing, heterogeneous computation • Statistics and Learning: Large scale clustering problems, compressive motion analysis, graphical inference • Embedded Systems: Board design, multi-processor interaction, interfaces, power/weight/form factor
  • 4. Mission Statement 4 To Teach Unmanned Vehicles to See as Humans Do Applications: • Absolute localization in GPS denied scenarios • Visual tracking odometry • Landmark detection and landmark-based navigation • Terrain mapping and change detection for IED disposal • Immersive visualization for situational awareness • Disaster response
  • 5. System Architecture 4 Server Nodes Producer Nodes Consumer Nodes
  • 6. System Architecture 5 Server Nodes Producer Nodes Consumer Nodes
  • 8. Producer Nodes 7 Computing Element: Embedded GPGPU processor ARM multi-core Heterogeneous architecture Optoelectronic Device: Active Stereo (IR) Passive Stereo Laser / TOF
  • 9. Producer Nodes 8 Key Features: • Visual Odometry ( < 3 mm ) • Mapping • Landmark Extraction • High Data Rates ( 9.2M pts / sec ) • Volumetric / Manifold Reconstruction • Source Compression for Uplink Compatibility: • Low power ( < 4.5 W ) • Low weight ( cell phone – battery ) • Low cost ( COTS sensors )
  • 10. Producer Nodes 9 Signal Processing • Fundamental task in vision-based algorithms • Most algorithms too slow even for desktops • Bilateral filters • Pyramid computation • Depth-RGB conversion • SIFT descriptors • Object recognition
  • 14. Producer Nodes 1 3 Visual Odometry • RGB-based • 30 FPS • Accuracy in mm • Compressive sensing solution • Range-based • 9.2M pts / sec • 90T comps / sec • Towards drift free?
  • 18. Producer Nodes 1 7 Can a mobile device map 3D environments to produce dense 3D data points rather than sparse landmarks? Dense Reconstruction • Landmark-based mapping (1960’s – present) – SOLVED • DARPA grand challenge • Efficient global alignment (1999 – 2005) – SOLVED • Multiple view geometry + dense reconstruction (1980’s – present) • Live dense reconstruction (2006 – present) • KinectFusion • Range-based SLAM • Dense Tracking and Mapping
  • 23. Producer Nodes 2 2 Patchwork Reconstructions • More memory efficient than volumes • Faster integration and tracking • Efficient caching interplay • Global refinement • Runs on embedded device
  • 25. Server Nodes 2 4 Global Refinement • Maps must be globally consistent • Dense reconstruction doesn’t allow for this refinement • Patchwork generalizes to this functionality • Inherently multi-core problem • ARM architecture for GraDeS
  • 27. Full Visualization 2 6 Custom Visualization • Key task: visualization • Interactivity (XBOX controller) • Full support for 3d / 2d streams • Event processing • Interoperability with GPGPU (CUDA) • Compression for consumer nodes • Fully scalable Rendering Pipeline • Custom rendering pipeline for Patchwork • Volumetric and Point Cloud • Raycasting with lighting sources
  • 29. Consumer Nodes 2 8 Our technology can produce maps at high speeds and unprecedented fidelities But.. What to do with this content? Visual Localization Situational Awareness • Localization a major problem in GPS- • Visualize mapping assets in real time denied scenarios from cell phones • “Urban canyons” • Coordinate with server and receive • Indoor environments compressed video stream • MAV / UGV coordination • Back-end models dynamics of • Existing solutions based mostly on adversaries state estimation • Extensible visualizer: new • Possible to query large maps for tags, models, data sources location?
  • 31. Feasible Path Infeasible Path
  • 33. Conclusion 3 2 Current trends in computer vision and robotics: • High performance computing • Live dense reconstruction • Range-based tracking and mapping Our architecture: • Producer nodes: • COTS sensors • Commodity computational unit • Dense tracking and mapping • Server nodes • Combine producer data into large maps • Serve consumer nodes • Consumer nodes • Visual absolute localization and remote visualization