The expansion of instruction-tuning data has enabled foundation language models to exhibit improved instruction adherence and superior performance across diverse downstream tasks. Semantically-rich 3D human motion is being progressively integrated with these foundation models to enhance multimodal understanding and cross-modal generation capabilities. However, the modality gap between human motion and text raises unresolved concerns about catastrophic forgetting during this integration. In addition, developing autoregressive-compatible pose representations that preserve generalizability across heterogeneous downstream tasks remains a critical technical barrier. To address these issues, we propose the Human Motion-Vision-Language Model (HMVLM), a unified framework based on the Mixture of Expert Low-Rank Adaption(MoE LoRA) strategy. The framework leverages the gating network to dynamically allocate LoRA expert weights based on the input prompt, enabling synchronized fine-tuning of multiple tasks. To mitigate catastrophic forgetting during instruction-tuning, we introduce a novel zero expert that preserves the pre-trained parameters for general linguistic tasks. For pose representation, we implement body-part-specific tokenization by partitioning the human body into different joint groups, enhancing the spatial resolution of the representation. Experiments show that our method effectively alleviates knowledge forgetting during instruction-tuning and achieves remarkable performance across diverse human motion downstream tasks.

Abstract 3D Human motion style transfer is a fundamental problem in computer graphic and animation processing. Existing AdaIN‐based methods necessitate datasets with balanced style distribution and content/style labels to train the clustered latent space. However, we may encounter a single unseen style example in practical scenarios, but not in sufficient quantity to constitute a style cluster for AdaIN‐based methods. Therefore, in this paper, we propose a novel two‐stage framework for few‐shot style transfer learning based on the diffusion model. Specifically, in the first stage, we pre‐train a diffusion‐based text‐to‐motion model as a generative prior so that it can cope with various content motion inputs. In the second stage, based on the single style example, we fine‐tune the pre‐trained diffusion model in a few‐shot manner to make it capable of style transfer. The key idea is regarding the reverse process of diffusion as a motion‐style translation process since the motion styles can be viewed as special motion variations. During the fine‐tuning for style transfer, a simple yet effective semantic‐guided style transfer loss coordinated with style example reconstruction loss is introduced to supervise the style transfer in CLIP semantic space. The qualitative and quantitative evaluations demonstrate that our method can achieve state‐of‐the‐art performance and has practical applications. The source code is available at https://github.com/hlcdyy/diffusion-based-motion-style-transfer .

Motion retargeting is a fundamental problem in computer graphics and computer vision. Existing approaches usually have many strict requirements, such as the source-target skeletons needing to have the same number of joints or share the same topology. To tackle this problem, we note that skeletons with different structure may have some common body parts despite the differences in joint numbers. Following this observation, we propose a novel, flexible motion retargeting framework. The key idea of our method is to regard the body part as the basic retargeting unit rather than directly retargeting the whole body motion. To enhance the spatial modeling capability of the motion encoder, we introduce a pose-aware attention network (PAN) in the motion encoding phase. The PAN is pose-aware since it can dynamically predict the joint weights within each body part based on the input pose, and then construct a shared latent space for each body part by feature pooling. Extensive experiments show that our approach can generate better motion retargeting results both qualitatively and quantitatively than state-of-the-art methods. Moreover, we also show that our framework can generate reasonable results even for a more challenging retargeting scenario, like retargeting between bipedal and quadrupedal skeletons because of the body part retargeting strategy and PAN.

3D human pose estimation has been a long-standing challenge in computer vision and graphics, where multi-view methods have significantly progressed but are limited by the tedious calibration processes. Existing multi-view methods are restricted to fixed camera pose and therefore lack generalization ability. This paper presents a novel Probabilistic Triangulation module that can be embedded in a calibrated 3D human pose estimation method, generalizing it to uncalibration scenes. The key idea is to use a probability distribution to model the camera pose and iteratively update the distribution from 2D features instead of using camera pose. Specifically, We maintain a camera pose distribution and then iteratively update this distribution by computing the posterior probability of the camera pose through Monte Carlo sampling. This way, the gradients can be directly back-propagated from the 3D pose estimation to the 2D heatmap, enabling end-to-end training. Extensive experiments on Human3.6M and CMU Panoptic demonstrate that our method outperforms other uncalibration methods and achieves comparable results with state-of-the-art calibration methods. Thus, our method achieves a trade-off between estimation accuracy and generalizability. Our code is in https://github.com/bymaths/probabilistictriangulation

Generating 3D human motion based on textual descriptions has been a research focus in recent years. It requires the generated motion to be diverse, natural, and conform to the textual description. Due to the complex spatio-temporal nature of human motion and the difficulty in learning the cross-modal relationship between text and motion, text-driven motion generation is still a challenging problem. To address these issues, we propose AttT2M, a two-stage method with multi-perspective attention mechanism: body-part attention and global-local motion-text attention. The former focuses on the motion embedding perspective, which means introducing a body-part spatio-temporal encoder into VQ-VAE to learn a more expressive discrete latent space. The latter is from the cross-modal perspective, which is used to learn the sentence-level and word-level motion-text cross-modal relationship. The text-driven motion is finally generated with a generative transformer. Extensive experiments conducted on HumanML3D and KIT-ML demonstrate that our method outperforms the current state-of-the-art works in terms of qualitative and quantitative evaluation, and achieve fine-grained synthesis and action2motion. Our code is in https://github.com/ZcyMonkey/AttT2M.

Predicting future motion based on historical motion sequence is a fundamental problem in computer vision, and it has wide applications in autonomous driving and robotics. Some recent works have shown that Graph Convolutional Networks(GCN) are instrumental in modeling the relationship between different joints. However, considering the variants and diverse action types in human motion data, the cross-dependency of the spatio-temporal relationships will be difficult to depict due to the decoupled modeling strategy, which may also exacerbate the problem of insufficient generalization. Therefore, we propose the Spatio-Temporal Gating-Adjacency GCN(GAGCN) to learn the complex spatio-temporal dependencies over diverse action types. Specifically, we adopt gating networks to enhance the generalization of GCN via the trainable adaptive adjacency matrix obtained by blending the candidate spatio-temporal adjacency matrices. Moreover, GAGCN addresses the cross-dependency of space and time by balancing the weights of spatio-temporal modeling and fusing the decoupled spatio-temporal features. Extensive experiments on Human 3.6M, AMASS, and 3DPW demonstrate that GAGCN achieves state-of-the-art performance in both short-term and long-term predictions.

Abstract Animating an avatar that reflects a user's action in the VR world enables natural interactions with the virtual environment. It has the potential to allow remote users to communicate and collaborate in a way as if they met in person. However, a typical VR system provides only a very sparse set of up to three positional sensors, including a head‐mounted display (HMD) and Optionally two hand‐held controllers, making the estimation of the user's full‐body movement a difficult problem. In this work, we present a data‐driven physics‐based method for predicting the realistic full‐body movement of the user according to the transformations of these VR trackers and simulating an avatar character to mimic such user actions in the virtual world in realtime. We train our system using reinforcement learning with carefully designed pretraining processes to ensure the success of the training and the quality of the simulation. We demonstrate the effectiveness of the method with an extensive set of examples.

3D human pose estimation is a fundamental problem in artificial intelligence, and it has wide applications in AR/VR, HCI and robotics. However, human pose estimation from point clouds still suffers from noisy points and estimated jittery artifacts because of handcrafted-based point cloud sampling and single-frame-based estimation strategies. In this paper, we present a new perspective on the 3D human pose estimation method from point cloud sequences. To sample effective point clouds from input, we design a differentiable point cloud sampling method built on density-guided attention mechanism. To avoid the jitter caused by previous 3D human pose estimation problems, we adopt temporal information to obtain more stable results. Experiments on the ITOP dataset and the NTU-RGBD dataset demonstrate that all of our contributed components are effective, and our method can achieve state-of-the-art performance.

Point clouds-based 3D human pose estimation that aims to recover the 3D locations of human skeleton joints plays an important role in many AR/VR applications. The success of existing methods is generally built upon large scale data annotated with 3D human joints. However, it is a labor-intensive and error-prone process to annotate 3D human joints from input depth images or point clouds, due to the self-occlusion between body parts as well as the tedious annotation process on 3D point clouds. Meanwhile, it is easier to construct human pose datasets with 2D human joint annotations on depth images. To address this problem, we present a weakly supervised adversarial learning framework for 3D human pose estimation from point clouds. Compared to existing 3D human pose estimation methods from depth images or point clouds, we exploit both the weakly supervised data with only annotations of 2D human joints and fully supervised data with annotations of 3D human joints. In order to relieve the human pose ambiguity due to weak supervision, we adopt adversarial learning to ensure the recovered human pose is valid. Instead of using either 2D or 3D representations of depth images in previous methods, we exploit both point clouds and the input depth image. We adopt 2D CNN to extract 2D human joints from the input depth image, 2D human joints aid us in obtaining the initial 3D human joints and selecting effective sampling points that could reduce the computation cost of 3D human pose regression using point clouds network. The used point clouds network can narrow down the domain gap between the network input i.e. point clouds and 3D joints. Thanks to weakly supervised adversarial learning framework, our method can achieve accurate 3D human pose from point clouds. Experiments on the ITOP dataset and EVAL dataset demonstrate that our method can achieve state-of-the-art performance efficiently.