Tracking loss in model-based pose estimation
Hi, thanks for this amazing work! I recently tried to use FoundationPose to track the object from RGB-D video recorded with my iPhone and Apple ARKit. However, I noticed that the FP can nicely track the object at the very beginning, then when my hand start to contact with the object, the tracking gradually loss. I attached the foundationpose the frame where the tracking start to loss here, could you please help me check why this tracking loss happens?
previously when I was recording data with RealSense d455 camera, everything goes well, FoundationPose can handle this level of occlusion nicely.
Best regards, Hilbert
Hey how long did Foundation pose take to train on this specific object ? I am trying to implement the DOPE algorithm but it takes too long to train .