Hi Wikitude team,
We are happy to see that the object recognition and tracking has greatly improved in the last version of the SDK.
The new technique using high fidelity photos to create an object target is impressive and allow to create very detailed object targets. But sometimes, these object targets could be too detailed and lead to long loading times of wto files. Also, when we have an app that only tries to recognise objects (and not to project any AR on it), we realised that all these details are not necessary. In this case, we end up taking fewer photos of an object, in lower resolutions, in order to create smaller .wto files and reduce loading time. But this process is not ideal, as it is difficult to predict what would be the amount of details of the target by selections the photos. Also sometimes we end up object targets with not enough details.
We think it would be great if we could have an option to export lighter version of .wto, where only the most significant points of an object would be selected (for recognition in particular). Also these .wto files could maybe contains more than 10 targets ?
This would help to present a rich AR experience, with fast loading times, on "older" phones.
Hope you can give this idea some thoughts,
Any comment on this idea @wikitude ?
thanks for sharing your idea how to improve our service, so that it better meets your requirements. If I'm correctly reading this you are struggling primarily with the size of the generated wto file? How big is the wto file getting and what are the phones you see issues with?
Additionally, you propose a kind of recognition-only mode, that would just tell you what object has been found, but not start place any augmentation in it nor do live tracking.
We don't have any immediate plans on the roadmap to address this, but will bring it forward to the team to see if there are any work-arounds