3D 3D Pos ose e Estimat ation on and and Mod odel el Ret - PowerPoint PPT Presentation

3D 3D Pos ose e Estimat ation on and and Mod odel el Ret Retriev eval al in n the he Wild Vincent Lepetit ENPC ParisTech & TU Graz

H-O3 O3D : Ha Hand+ d+Obj Object Dataset Da 3D pos 3D ose, e, 3D 3D mod odel el retri re rieval in the wild 2

H-O3 O3D : Ha Hand+ d+Obj Object Dataset Da 3D pos 3D ose, e, 3D 3D mod odel el retri re rieval in the wild 3

3D Pose Estimation of Rigid Objects BB8 BB8: A A Scalable, Ac Accurate, Ro Robust to Partial Occlusion Method for Predicting the 3D Poses of of Chal halleng enging ng Object ects without hout Using ng Dep epth . Mahdi Rad and Vincent Lepetit. ICCV 2017. 4

<latexit sha1_base64="ihyU/UrEqlV9uKtS1GtUl3ZCPFk=">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</latexit> <latexit sha1_base64="pNG6kId1jlsLq98/nc1sZl0PXfc=">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</latexit> 3D Pose Estimation from Correspondences M • Predicting 2D locations from an image is an easier regression task; • m We do not need a representation of the 3D m 1 m 3 rotation; m 4 • We do not need to balance the rotation and m 2 the translation. We can compute the 3D pose from these 2D locations. 5 Camera center

3D Pose Retrieval for Object Categories 3D 3D Pose ose Est stim imat ation ion and and 3D 3D Mod odel el Ret Retrieval rieval for or Object jects s in in the he Wild ild . Alexander Grabner, Peter M. Roth, and Vincent Lepetit. CVPR 2018. 6

3D Pose Retrieval for Object Categories 2d bounding boxes 7

3D Pose Retrieval for Object Categories pose predictor 3D pose ? 8

3D Pose Retrieval for Object Categories 2D 3D pose of the reprojections object’s bounding box P n P network (length, width, height) of height object’s bounding box width length 9

3D Model Retrieval for Object Categories Locat Location ion Field Field Descrip escriptors: ors: Sing ingle le Imag age e 3D 3D Mod odel el Ret Retrieval rieval in in the he Wild ild. Alexander Grabner, Peter M. Roth, and Vincent Lepetit. 3DV 2019. 10

<latexit sha1_base64="0GjcsQugI5c5fp4RGqChCknagkE=">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</latexit> <latexit sha1_base64="0GjcsQugI5c5fp4RGqChCknagkE=">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</latexit> 3D Model Retrieval for Object Categories ⇣ . ⌘ ⇣ . ⌘ pose invariant . . descriptors . . ShapeNet [Chang et al, 2015] 11

Location Fields 12