Visual-SLAM

EEC 193AB Independent Senior Design Project 2018-2019 @ UC-Davis
Final Project Report

Group Members

• Kolin Guo
• Jeff Lai
• Wenda Xu

Teaching Assistants

• Teja Aluru
• Adam Jones
• Minh Truong

Instructions on Setting up This Repository

Prerequisites

The list of prerequisites for building and running this repository is described below.

• GNU/Linux x86_64 with kernel version > 3.10
• Docker >= 1.12
• NVIDIA GPU with Architecture > Fermi (2.1)
• NVIDIA drivers >= 418.39
  (Required for CUDA 10.1.105, can change Dockerfile base image to downgrade CUDA version if NVIDIA driver doesn't satisfy. Driver version required for different CUDA versions can be found in Table 1. Other base images can be found here)
• nvidia-docker2

Setup Instructions

The setup process is separated into two shell scripts: setup.sh and build.sh. Both have some command-line arguments that you can specify to config them. See the printed out usage for details.
setup.sh might take more than an hour to finish depending on your CPU power and network environment. Please be patient and wait until it finishes completely.

1. Setup Docker Environment.
   bash ./setup.sh
   If you need sudo permission to run nvidia-docker, run sudo -s before running setup.sh.
   You should be greeted by the Docker container orbslam2py when this script finishes. The working directory is /root and the repo is mounted at /root/Visual-SLAM.

2. Build the C++11 and Python ORBSLAM2 Implementation.
   cd /root/Visual-SLAM && bash -i ./build.sh && source ~/.bashrc
   This script should be running in a Docker container to gain access to dependencies. It needs to be run interactively with -i option so that it can do source ~/.bashrc.
   Instructions on how to run C++11 and Python implementation will be printed when this script finishes.
   If failed to open X display, run xhost + on local computer.

Notes

• The testing/benchmarking environment of the original ORB_SLAM2 C++ implementation is the Benchmark branch. We avoid benchmarking in the master branch to keep it tidy. All benchmark results will be summarized and uploaded to master branch.