Meta tags:
description=Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses;
keywords=Pose Transfer, Deformation Transfer, Hybrid Representation, Jacobian Fields;

Headings (most frequently used words):

qualitative, using, results, neural, pose, representation, learning, for, generating, and, transferring, non, rigid, object, poses, abstract, method, deformingthings4d, animals, comparisons, smpl, human, body, shapes, adobe, mixamo, citation,

Text of the page (most frequently used words):
the (33), pose (23), and (18), #object (10), poses (10), for (8), shapes (7), variations (6), #representation (5), learning (5), our (5), using (5), from (5), applier (5), target (5), shape (5), with (5), based (4), neural (4), transferring (4), are (4), method (4), information (4), generating (3), non (3), rigid (3), qualitative (3), results (3), representations (3), extracted (3), implicit (3), diffusion (3), model (3), objects (3), generation (3), transfer (3), transferred (3), website (2), source (2), code (2), yoo (2), seungwoo (2), koo (2), juil (2), yeo (2), kyeongmin (2), sung (2), minhyuk (2), human (2), keypoint (2), hybrid (2), jacobian (2), fields (2), these (2), mapped (2), extractor (2), deformation (2), field (2), various (2), training (2), trained (2), self (2), supervised (2), manner (2), latent (2), diverse (2), propose (2), novel (2), deformable (2), other (2), identities (2), level (2), first (2), better (2), cascaded (2), left (2), right (2), rows (2), thank, kindly, open, sourcing, keunhong, park, nerfies, this, licensed, under, creative, commons, attribution, sharealike, international, license, inproceedings, yoo2024neuralpose, title, author, booktitle, neurips, year, 2024, please, consider, citing, work, you, find, useful, citation, adobe, mixamo, comparisons, smpl, body, deformingthings4d, animals, framework, extracts, back, conditioned, acts, including, those, unseen, during, refinement, module, alpha, positioned, between, leveraging, template, compactness, facilitates, enabling, through, generative, modeling, space, overview, which, specializes, disentangling, identity, facilitating, properties, enables, diversity, both, single, does, not, require, explicit, parameterization, such, skeletons, joints, point, correspondence, supervision, design, represent, learn, distill, geometry, output, intrinsic, mesh, property, face, once, fine, tuned, describe, also, used, train, enable, experiments, deformthings4d, datasets, demonstrate, state, art, performance, ability, generate, deformed, abstract, motion, sequence, variation, proposed, third, second, fourth, preserving, intricate, details, like, horns, antlers, new, sampled, bunny, last, column, animal, teaser, arxiv, kaist,

Text of the page (random words):
neural pose representation learning for generating and transferring non rigid object poses neural pose representation learning for generating and transferring non rigid object poses seungwoo yoo juil koo kyeongmin yeo minhyuk sung kaist arxiv code teaser results of motion sequence transfer left and shape variation generation right using the proposed neural pose representation on the left poses from source shapes first and third rows are transferred to target shapes second and fourth rows preserving intricate details like horns and antlers on the right new poses sampled from a cascaded diffusion model trained with shape variations of the bunny last column are transferred to other animal shapes abstract we propose a novel method for learning representations of poses for 3d deformable objects which specializes in 1 disentangling pose information from the object s identity 2 facilitating the learning of pose variations and 3 transferring pose information to other object identities based on these properties our method enables the generation of 3d deformable objects with diversity in both identities and poses using variations of a single object it does not require explicit shape parameterization such as skeletons or joints point level or shape level correspondence supervision or variations of the target object for pose transfer we first design the pose extractor to represent the pose as a keypoint based hybrid representation and the pose applier to learn an implicit deformation field to better distill pose information from the object s geometry we propose the implicit pose applier to output an intrinsic mesh property the face jacobian once the extracted pose information is transferred to the target object the pose applier is fine tuned in a self supervised manner to better describe the target object s shapes with pose variations the extracted poses are also used to train a cascaded diffusion model to enable the generation of novel poses our experiments with the deformthings4d and human datasets demonstrate state of the art performance in pose transfer and the ability to generate diverse deformed shapes with various objects and poses method method overview our framework extracts keypoint based hybrid pose representations from jacobian fields these fields are mapped by the pose extractor g and mapped back by the pose applier h the pose applier conditioned on the extracted pose acts as an implicit deformation field for various shapes including those unseen during training a refinement module alpha positioned between g and h is trained in a self supervised manner leveraging the target s template shape the compactness of our latent representations facilitates the training of a diffusion model enabling diverse pose variations through generative modeling in the latent space qualitative results using deformingthings4d animals qualitative comparisons using smpl human body shapes qualitative results using adobe mixamo citation please consider citing our work if you find it useful inproceedings yoo2024neuralpose title neural pose representation learning for generating and transferring non rigid object poses author yoo seungwoo and koo juil and yeo kyeongmin and sung minhyuk booktitle neurips year 2024 this website is licensed under a creative commons attribution sharealike 4 0 international license the website is based on nerfies we thank keunhong park for kindly open sourcing the source code