Detect occluded object in image and get orientation without train using CAD model with template-pose
Detect occluded object in image and get orientation without train using CAD model with template-pose
Templates for 3D Object Pose Estimation Revisited: Generalization to New Objects and Robustness to Occlusions
arXiv paper abstract https://arxiv.org/abs/2203.17234v1
arXiv PDF paper https://arxiv.org/pdf/2203.17234v1.pdf
… present a method that can recognize new objects and estimate their 3D pose in RGB images even under partial occlusions.
… requires neither a training … on these objects nor real images depicting them, only their CAD models.
It relies on a small set of training objects to learn local object representations, which allow us to locally match the input image to a set of “templates”, rendered images of the CAD models for the new objects.
In contrast with the state-of-the-art methods, the new objects on which … method is applied can be very different from the training objects.
… first to show generalization without retraining on the LINEMOD and Occlusion-LINEMOD datasets.
… outperform the state-of-the-art template matching methods on the LINEMOD, Occlusion-LINEMOD and T-LESS datasets. …
Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website
LinkedIn https://www.linkedin.com/in/morris-lee-47877b7b
