Expanding the predefined dictionaries

[EceChaik]

Hi everybody,

I was wondering if the MIP_36h12 dictionary can be added to opencv_contrib/aruco

[bign8]

Is this still wanted or in the works?

Looks like the ArUco authors highly recommend the use of the use of the MIP_36h12 library.

We ... recommend using ARUCO_MIP_36h12 because it is more robust to errors.
Source

After some poking around over there, I found the dictionary in src/dictionary.cpp#L159.

If this is still wanted or open to be contributed, the data has to be converted into a format that the predefined dictionaries expect:

Dictionaries are stored as a list of bytes in its four rotations
On each rotation, the marker is divided in bytes assuming a row-major order
This format allows a faster marker identification.
For a dictionary composed by M markers of NxN bits, the structure dimensions should be:
const char name[nMarkers][4rotations][nBytes], or more specifically:
const char name[M][4][ceil(NxN/8)]
The element [i][j][k] represents the k-th byte of the i-th marker in the dictionary
in its j-th rotation.
Each rotation implies a 90 degree rotation of the marker in anticlockwise direction.
Source: modules/aruco/src/predefined_dictionaries.cpp

Any details around why this issue was closed? Is there a simple fix that I'm not seeing?

Thanks,

• an OpenCV (and ArUco) newbie

[asmorkalov]

related: #3200

[Em607867]

Is this still wanted or in the works?

Looks like the ArUco authors highly recommend the use of the use of the MIP_36h12 library.

We ... recommend using ARUCO_MIP_36h12 because it is more robust to errors.
Source

After some poking around over there, I found the dictionary in src/dictionary.cpp#L159.

If this is still wanted or open to be contributed, the data has to be converted into a format that the predefined dictionaries expect:

Dictionaries are stored as a list of bytes in its four rotations
On each rotation, the marker is divided in bytes assuming a row-major order
This format allows a faster marker identification.
For a dictionary composed by M markers of NxN bits, the structure dimensions should be:
const char name[nMarkers][4rotations][nBytes], or more specifically:
const char name[M][4][ceil(NxN/8)]
The element [i][j][k] represents the k-th byte of the i-th marker in the dictionary
in its j-th rotation.
Each rotation implies a 90 degree rotation of the marker in anticlockwise direction.
Source: modules/aruco/src/predefined_dictionaries.cpp

Any details around why this issue was closed? Is there a simple fix that I'm not seeing?

Thanks,

• an OpenCV (and ArUco) newbie

Hi,

Were you able to resolve this issue? I am trying to detect markers in the same dictionary

[asmorkalov]

Python bindings for dicts: #3229

[AleksandrPanov]

Add dictionary bindings #3229 was merged