ChallenCap: Monocular 3D Capture of Challenging Human Performances using Multi-Modal References

Abstract

Capturing challenging human motions is critical for numerous applications, but it suffers from complex motion patterns and severe self-occlusion under the monocular setting. In this paper, we propose ChallenCap -a template-based approach to capture challenging 3D human motions using a single RGB camera in a novel learning-and-optimization framework, with the aid of multi-modal references. We propose a hybrid motion inference stage with a generation network, which utilizes a temporal encoder-decoder to extract the motion details from the pair-wise sparse-view reference, as well as a motion discriminator to utilize the unpaired marker-based references to extract specific challenging motion characteristics in a data-driven manner. We further adopt a robust motion optimization stage to increase the tracking accuracy, by jointly utilizing the learned motion details from the supervised multi-modal references as well as the reliable motion hints from the input image reference. Extensive experiments on our new challenging motion dataset demonstrate the effectiveness and robustness of our approach to capture challenging human motions.

Recovering 3D Human Mesh from Monocular Images: A Survey

“…Besides, TCMR [114] applies two more GRUs to forecast additional temporal features for the current target pose from the past and future frames. Based on GRUs and a motion discriminator, ChallenCap [198] utilizes multi-modal and hybrid motion references to capture challenging human motions. Lee et al [147] consider the uncertainty-aware embedding and include optical flow information.…”

“…MPoser, an extension of VPoser [22] to temporal sequences, is based on sequential VAE. Inspired by VIBE, He et al [198] generate marker-based motion maps as input to a discriminator to obtain an adversarial motion prior. In HuMoR [158], the probability distribution of possible state transitions is formulated by a conditional variational autoencoder (CVAE).…”

NeuVV: Neural Volumetric Videos with Immersive Rendering and Editing

“…There are many real-time motion capture solutions available and we adopt the recent single camera technique [He et al 2021] for convenience. It is able to detect 21 key points of skeletons.…”

“…In this specific case, a user should not only be able to omnidirectionally watch the virtual trainer's moves but also compare their own moves with the trainer. In our implementation, we use a single camera motion capture solution [He et al 2021] that estimates 3D skeleton structures of users as they move. We also precompute the "ground truth" skeleton moves of the trainer, by first rendering a multiview video of whole body movements also using NeuVV and then conducting multi-view skeleton estimation.…”

Imposing Temporal Consistency on Deep Monocular Body Shape and Pose Estimation

“…To capture humans in challenging poses and motion, ChallenCap [7] uses a learning-and-optimization framework, which learns motion characteristics and is trained on "a new challenging human motion dataset with both unsynchronized marker-based and light-weight multiimage references" [7]. First, a noisy skeletal motion map is acquired, this is then processed in their temporal encoder-decoder and generation network HybridNet and their motion discriminator, which is followed by a robust motion optimization phase, which uses 2D keypoint and silhouette information from the video frames.…”