FLaME: Fast Lightweight Mesh Estimation
FLaME (Fast Lightweight Mesh Estimation) is a lightweight, CPU-only method
for dense online monocular depth estimation. Given a sequence of camera images
with known poses, FLaME is able to reconstruct dense 3D meshes of the
environment by posing the depth estimation problem as a variational optimization
over a Delaunay graph that can be solved at framerate, even on computationally
constrained platforms.
The flame
repository contains the source code for the core algorithm. It
should be input/output agnostic, so feel free to write an appropriate frontend
for your data. ROS bindings are available with the
associated flame_ros
repository, which also includes examples for running flame
on offline data.
NOTE: These instructions assume you are running Ubuntu 16.04 and are
interested in installing flame
only. See the installation instructions forflame_ros
if you also wish to build the ROS bindings as the process can be
streamlined using catkin_tools
.
Install apt
dependencies:
sudo apt-get install libboost-all-dev
Install OpenCV 3.2:
Unfortunately OpenCV 3.2 is not available through apt
on Ubuntu 16.04. If you
have ROS Kinetic installed on your system, you can simply source your ROS
installation as this version of OpenCV is packaged with ROS Kinetic. If you
don’t have ROS Kinetic installed, then you will need to install from
source. Please consult
the
OpenCV docs for
instructions.
mkdir -p dependencies/src
./scripts/eigen.sh ./dependencies/src ./dependencies
./scripts/sophus.sh ./dependencies/src ./dependencies
cp ./scripts/env.sh ./dependencies/
source ./dependencies/env.sh
4. Install `flame`:
```bash
cd flame
mkdir build
cd build
cmake -D CMAKE_INSTALL_PREFIX=path/to/install/directory ..
make install
See flame_ros
for ROS bindings and example usage.