This paper addresses the challenging task of reconstructing the poses of multiple individuals engaged in close interactions, captured by multiple calibrated cameras. The difficulty arises from the noisy or false 2D keypoint detections due to inter-person occlusion, the heavy ambiguity in associating keypoints to individuals due to the close interactions, and the scarcity of training data as collecting and annotating motion data in crowded scenes is resource-intensive. We introduce a novel system to address these challenges. Our system integrates a learning-based pose estimation component and its corresponding training and inference strategies. The pose estimation component takes multi-view 2D keypoint heatmaps as input and reconstructs the pose of each individual using a 3D conditional volumetric network. As the network doesn't need images as input, we can leverage known camera parameters from test scenes and a large quantity of existing motion capture data to synthesize massive training data that mimics the real data distribution in test scenes. Extensive experiments demonstrate that our approach significantly surpasses previous approaches in terms of pose accuracy and is generalizable across various camera setups and population sizes. The code is available on our project page: https://github.com/zju3dv/CloseMoCap.

Deterministic models dominate the field of talking facial land-mark generation by directly mapping speech signals to a certain lip-sync facial landmark sequence, which often suffer from regression to the mean face. In contrast, probability generative models are more beneficial to handle complex data space and generate diverse samples. In this work, we pro-pose a flow-based probabilistic network named TalkingFlow to generate natural talking facial landmark with head movements from speech data. It is implemented by a weighted multi-scale architecture to improve model representation capability and a conditional Temporal Convolutional Network module to fuse speech data. Extensive experiments results show that it can effectively generate diverse and natural facial landmark from speech data. All code will be made publicly available online.

This paper presents a simple yet effective approach for learning 6DoF object poses without real image annotations. Previous methods have attempted to train pose estimators on synthetic data, but they do not generalize well to real images due to the sim-to-real domain gap and produce inaccurate pose estimates. We find that, in most cases, the synthetically trained pose estimators are able to provide reasonable initialization for depth-based pose refinement methods which yield accurate pose estimates. Motivated by this, we propose a novel learning framework, which utilizes the accurate results of depth-based pose refinement methods to supervise the RGB-based pose estimator. Our method significantly outperforms previous self-supervised methods on several benchmarks. Even compared with fully-supervised methods that use real annotated data, we achieve competitive results without using any real annotation. The code is available at https://github.com/zju3dv/pvnet-depth-sup.

In this paper, we present an improved framework for the instance-aware semantic segmentation of road furniture in mobile laser scanning data. In our framework, we first detect road furniture from mobile laser scanning point clouds. Then we decompose the detected pieces of road furniture into poles and their attached components, and extract the instance information of the components with different features. Most importantly, we classify the components into different categories by combining a classifier and a probabilistic graphic model named DenseCRF, which is the major contribution of this paper. For the classification of the components using DenseCRF, the unary potentials and the pairwise potentials are first obtained. The unary potentials are obtained from the classifier which takes the instance information of components as the input. The pairwise potentials are calculated considering contextual relations between components. By utilising DenseCRF, the contextual consistency of components is preserved, and the performance is significantly improved compared to our previous work. We collect three datasets to test our framework, and compare the classification performances of six different classifiers with and without DenseCRF. The combination of random forest with DenseCRF outperforms the other methods and achieves high overall accuracies of 83.7%, 96.4% and 95.3% in these three datasets. Experimental results demonstrate that our framework reliably assigns both semantic information and instance information for mobile laser scanning point clouds of road furniture.