This is the repository for Seedmarkers - Embeddable Markers for Physical Objects, shape-independent topological markers that can be embedded in physical objects manufactured with common rapid-prototyping techniques.
Install opencv depending on your operating system (for ArUco opencv-contrib is required)
Install python3 dependencies: pip3 install -r requirements.txt
Install the requirements as stated above and run generate_example_shapes.sh in the generate directory.
PNG output can be found in seedmarker_debug_clear.
Import the generated DXF and extrude the marker. Emboss it in the object and orient this face towards the buildplate when printing. Configure "pause at Z" in your slicing software and let the printer pause and eject filament after the first layer. Change filament and continue.
For lasercutting post-process the DXF (conversion to SVG may be necessary) according to the needs of your lasercutters software package. Sometimes colouring the lines with a certain color is necessary, sometimes manual selection of polygons for engraving needs to be done.
See the paper for more details.
For 6-DoF pose estimation camera calibration data is required. You can obtain that with openCV and copy the matrices to a JSON file (see detect/calibration_pixel3a.json for the structure).
Run detect/detection_video.sh for a short demo.
detect/detection.py --help will print an overview of the command line arguments.
The generator requires an DXF-file that either contains circles or Polylines. Arcs, splines or other primitives are not supported (polyline-bulges are supported). Note: there are a few tricks to force Polyline output. Fusion360 will use Polylines if fillets are used at the end of arc-segments.
python3 generate.py example_shapes/circle.dxf 01232323423 --output circle_marker.dxf
If an orientation vector (0-360 deg) and a ReacTIVision amoeba graph is supplied (with or without the leading w/b), the generator will try to arrange the leaves in a way that results in a clear separation between white and black circles. Note that this is more error-prone on complex graphs and outlines. Some geometries will work well, some will terminate early.
See generate/generate_reactivision.sh
PNGs that are generated are vertically flipped (DXF coordinate system lower left corner, PIL coordinate system upper left corner). If the marker is imprinted using the DXF on the bottom of a 3D printed part the orientation is correct. If the PNG is using for inkjet-printing, vertical mirroring is required.
Use a graph that has at least 4 unique subtrees with only a single leaf on the last layer. See generate/SeedmarkerSpeaker_no_cutout3_marker_print.txt for an example. The detector requires a OpenCV-style calibration matrix in a JSON-file (see detect/iphonexs_calibration.json).
Depending on the size of the graph or the area of the provided geometry the generator may fail to find a solution. Try decreasing the size of the graph or the minimal line width (append --line-width=1.0)
If reading the DXF aborts because no valid geometries could be found, your file may contain splines. Please see the DXF paragraph.
You can find the open-access paper in the ACM library: Seedmakers.pdf
If you want to use and cite this work:
@inproceedings{10.1145/3430524.3440645,
author = {Getschmann, Christopher and Echtler, Florian},
title = {Seedmarkers: Embeddable Markers for Physical Objects},
year = {2021},
url = {https://doi.org/10.1145/3430524.3440645},
doi = {10.1145/3430524.3440645},
booktitle = {Fifteenth International Conference on Tangible, Embedded, and Embodied Interaction},
location = {Salzburg, Austria},
series = {TEI '21}
}