ElasticFusion Dockerfile

Dockerfile for use of ElasticFusion with RealSense

Requirements

• Docker
  • from: https://docs.docker.com/get-docker/
• NVIDIA Container Toolkit
  • from: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html

My environment (ref.)

• Ubuntu 20.04
• CUDA 11.2 (host)

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 460.32.03    Driver Version: 460.32.03    CUDA Version: 11.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  GeForce GTX 1080    On   | 00000000:01:00.0  On |                  N/A |
| 46%   52C    P2    67W / 180W |   3900MiB /  8116MiB |     26%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

Docker build and run

$ docker build -t <image_name> ./docker
$ xhost local:
$ ./opendocker.sh <image_name>
$ xhost -local:

Run with RealSense

I tested only with RealSense D435.

$ ElasticFusion
# data is saved as `/opt/ElasticFusion/GUIlive.ply`

Run with sample data

$ wget http://www.doc.ic.ac.uk/~sleutene/datasets/elasticfusion/dyson_lab.klg -P ./workspace
# in container
$ ElasticFusion -l dyson_lab.klg

Visualize result

$ pipenv sync
$ pipenv shell
$ python visualize.py --ply <path/to/.ply>