prime-slam / glass-detection-dockers

This repository comprises glass segmentation algorithms and corresponding dockerfiles.

prime-slam

SLAM system using various types of landmarks. README will be supplemented.

Building Docker Image

docker build -t prime-slam .

Running Docker Container

To run the container use the following command:

docker run --rm -v <DATA_PATH>:/data prime-slam [OPTIONAL_ARGS]

The following [OPTIONAL_ARGS] can be used:

optional arguments:
  -h, --help            show this help message and exit
  --data PATH, -d PATH  path to data (default: data/)
  --data-format STR, -D STR
                        data format: tum, icl, icl_tum (default: icl_tum)
  --save-cloud BOOL, -s BOOL
                        save resulting cloud (default: True)
  --cloud-save-path PATH, -S PATH
                        path to the saved cloud (default: resulting_cloud.pcd)
  --verbose BOOL, -v BOOL
                        print metrics (default: True)

Data formats

icl

/data
├── scene_0.depth — depth in ICL format
├── scene_0.png — image
├── scene_0.txt — camera parameters
├── scene_1.depth
├── scene_1.png
├── scene_1.txt
...
├── scene_N.depth
├── scene_N.png
└── scene_N.txt

File names may differ from those shown above. Additional information can be found here.

tum

/data
├── /rgb — depth in ICL format
    ├── image_0.png
    ...
    └── image_N.png
├── /depth
    ├── depth_0.png
    ...
    └── depth_M.png
└── groundtruth.txt — gt poses in TUM format

Additional information can be found here.

icl_tum

ICL data presented in tum format.