Feature webarkit trackers and beyond -> PlanarTracker integration#16
Feature webarkit trackers and beyond -> PlanarTracker integration#16
Conversation
WebARKitOrbTracker classes. Now only WebARKitTracker is necessary. - fix for issue webarkit/webarkit-testing#11
|
I have solved the issue #11 with these commits 5237407 and 323b6df |
classes - code under testing phase, but no errors while compiling
|
I started to integrate the PlanarTracker code from artoolkitx into the WebARKitTracker class. If it works well on desktop and Mobile native devices, it will works also with Emscripten. There are no reason that it can not works. Probably theere will be differences in performances until we'll not add simd support during compilation of OpenCV libs and during the compilation of the final stage. But this this is another chapter of the development. |
- HomographyInfo, TrackableInfo, TrackedPoint, TrackingPointSelector, - more configs - Points and GetHomograpyInliers
UpdateTrackableHomography from artoolkitx
GetVerticesFromPoint GetVerticesFromTopCorner GetTemplateRoi IsRoiValidForFrame InflateRoi
- code can be compiled but can not be used yet
GetTrackablePose functions
|
I think will open another PR starting from commit 03b8f4e before applying code from PlanarTracker (artoolkitx). This last code is not stable and is not guaranteed. |
| } | ||
|
|
||
| bool CanDetectNewFeatures() { return (_currentlyTrackedMarkers < _maxNumberOfMarkersToTrack); } | ||
| //bool CanDetectNewFeatures() { return (_currentlyTrackedMarkers < _maxNumberOfMarkersToTrack); } |
There was a problem hiding this comment.
I had to override this because _maxNumberOfMarkersToTrack result = 0 !! so the boolean comparison is resulting always false and the detection routine never start. _maxNumberOfMarkersToTrack is initialized to 1, but if i print is value is always 0, so there is a bug somewhere....
cv::Mat as reference instead
f301840 to
6c3d0c6
Compare
|
Now the code start to detect and track, i can see in the console the message |
I will discuss here changes and improves in the WebARKitTracker code.
This is the follow up of the development started on my fork see this PR kalwalt#4
First i need to solve issue webarkit/webarkit-testing#11 this is a must.
My findings: probably the issue arise because the unique_ptr share the same track method.
Second i think the abstract class is not needed to derive the other classes (WebARKitOrbTracker and WebARKitAkazeTracker classes) i can simplify the code implementing a new method setDetectorType to switch the detector (Orb or Akaze). Will be nice also to implemement a Pimpl pattern for the WebARKitTracker class.