A standalone portable 3D reconstruction system with Intel realsense and Jetson Nano

• Standalone 3D reconstruction system
• Instant 3D reconstruction
• Metrically accurate
• Low power consuming
• Offline and online (both)
• Cost effective
• Modular approach
• Deep learning integration (3D deep learning for registration)

Hardware built:

Reconstructed results:

Jetscan is a portable 3D dense reconstruction system based on

1. Jetson nano by NVIDIA
2. Intel Realsense D400 series depth camera
3. Open3D by INTEL ISL
4. theNded's CUDA implementation

• For setting up the software stack, follow instructions in

  Sofware_O3D - Software setup

• Complete software repo is here (forked from Open3D)

  https://github.com/devshank3/Open3D-for-Jetson

  its built on theNded's CUDA implementation

Follow instructions in

1. Hardware_electronics - Electronics setup

2. Hardware_casing - Case/Closure setup

Follow https://github.com/devshank3/JetScan/blob/master/Software_O3D/README.md for Software Setup

• Step 1 : Go to Open3D-for-Jetson/Reconstruction_Jetscan_Hybrid/

• Step 2 : Edit Main_gui.py

  set your system executable path to Run_MakeFragments,ViewPoseGraph,Run_IntegrateScene

  eg.

   os.system('/home/shank/Projects/3D_processing/jetscan/jetscan-final-edit/Open3D-for-Jetson/build/bin/examples/./Run_MakeFragments')
  
                                                          ⇩⇩⇩ to ⇩⇩⇩                       
  
   os.system('your_system _path/Open3D-for-Jetson/build/bin/examples/./Run_MakeFragments')
  

• Step 3: Run the UI (in virtual env if setup)

  $ python3 Main_gui.py

• Step 4 : Start the Reconstruction

  Capture RGB-D

  Captures the RGB-D sequence

  Reconstruct Scene

  The hybrid reconstruction system starts with the RGB-D sequence recorded

  View

  View the latest reconstructed 3D Model

If interested pls contact me @ [email protected]

Under the MIT License (MIT)

Copyright (c) 2020 Shankar Hariharan (aka devshank3)