This paper presents an inverse kinematic optimization layer (IKOL) for 3D human pose and shape estimation that leverages the strength of both optimization- and regression-based methods within an end-to-end framework. IKOL involves a nonconvex optimization that establishes an implicit mapping from an image's 3D keypoints and body shapes to the relative body-part rotations. The 3D keypoints and the body shapes are the inputs and the relative body-part rotations are the solutions. However, this procedure is implicit and hard to make differentiable. So, to overcome this issue, we designed a Gauss-Newton differentiation (GN-Diff) procedure to differentiate IKOL. GN-Diff iteratively linearizes the nonconvex objective function to obtain Gauss-Newton directions with closed form solutions. Then, an automatic differentiation procedure is directly applied to generate a Jacobian matrix for end-to-end training. Notably, the GN-Diff procedure works fast because it does not rely on a time-consuming implicit differentiation procedure. The twist rotation and shape parameters are learned from the neural networks and, as a result, IKOL has a much lower computational overhead than most existing optimization-based methods. Additionally, compared to existing regression-based methods, IKOL provides a more accurate mesh-image correspondence. This is because it iteratively reduces the distance between the keypoints and also enhances the reliability of the pose structures. Extensive experiments demonstrate the superiority of our proposed framework over a wide range of 3D human pose and shape estimation methods.
Contrastive language-image pre-training (CLIP) achieves promising results in 2D zero-shot and few-shot learning. Despite the impressive performance in 2D tasks, applying CLIP to help the learning in 3D scene understanding has yet to be explored. In this paper, we make the first attempt to investigate how CLIP knowledge benefits 3D scene understanding. To this end, we propose CLIP2Scene, a simple yet effective framework that transfers CLIP knowledge from 2D image-text pre-trained models to a 3D point cloud network. We show that the pre-trained 3D network yields impressive performance on various downstream tasks, i.e., annotation-free and fine-tuning with labelled data for semantic segmentation. Specifically, built upon CLIP, we design a Semantic-driven Cross-modal Contrastive Learning framework that pre-trains a 3D network via semantic and spatial-temporal consistency regularization. For semantic consistency regularization, we first leverage CLIP's text semantics to select the positive and negative point samples and then employ the contrastive loss to train the 3D network. In terms of spatial-temporal consistency regularization, we force the consistency between the temporally coherent point cloud features and their corresponding image features. We conduct experiments on the nuScenes and SemanticKITTI datasets. For the first time, our pre-trained network achieves annotation-free 3D semantic segmentation with 20.8\% mIoU. When fine-tuned with 1\% or 100\% labelled data, our method significantly outperforms other self-supervised methods, with improvements of 8\% and 1\% mIoU, respectively. Furthermore, we demonstrate its generalization capability for handling cross-domain datasets.
Domain adaptation for Cross-LiDAR 3D detection is challenging due to the large gap on the raw data representation with disparate point densities and point arrangements. By exploring domain-invariant 3D geometric characteristics and motion patterns, we present an unsupervised domain adaptation method that overcomes above difficulties. First, we propose the Spatial Geometry Alignment module to extract similar 3D shape geometric features of the same object class to align two domains, while eliminating the effect of distinct point distributions. Second, we present Temporal Motion Alignment module to utilize motion features in sequential frames of data to match two domains. Prototypes generated from two modules are incorporated into the pseudo-label reweighting procedure and contribute to our effective self-training framework for the target domain. Extensive experiments show that our method achieves state-of-the-art performance on cross-device datasets, especially for the datasets with large gaps captured by mechanical scanning LiDARs and solid-state LiDARs in various scenes. Project homepage is at https://github.com/4DVLab/CL3D.git
Depth estimation is usually ill-posed and ambiguous for monocular camera-based 3D multi-person pose estimation. Since LiDAR can capture accurate depth information in long-range scenes, it can benefit both the global localization of individuals and the 3D pose estimation by providing rich geometry features. Motivated by this, we propose a monocular camera and single LiDAR-based method for 3D multi-person pose estimation in large-scale scenes, which is easy to deploy and insensitive to light. Specifically, we design an effective fusion strategy to take advantage of multi-modal input data, including images and point cloud, and make full use of temporal information to guide the network to learn natural and coherent human motions. Without relying on any 3D pose annotations, our method exploits the inherent geometry constraints of point cloud for self-supervision and utilizes 2D keypoints on images for weak supervision. Extensive experiments on public datasets and our newly collected dataset demonstrate the superiority and generalization capability of our proposed method.
LiDAR can capture accurate depth information in large-scale scenarios without the effect of light conditions, and the captured point cloud contains gait-related 3D geometric properties and dynamic motion characteristics. We make the first attempt to leverage LiDAR to remedy the limitation of view-dependent and light-sensitive camera for more robust and accurate gait recognition. In this paper, we propose a LiDAR-camera-based gait recognition method with an effective multi-modal feature fusion strategy, which fully exploits advantages of both point clouds and images. In particular, we propose a new in-the-wild gait dataset, LiCamGait, involving multi-modal visual data and diverse 2D/3D representations. Our method achieves state-of-the-art performance on the new dataset. Code and dataset will be released when this paper is published.
HD map reconstruction is crucial for autonomous driving. LiDAR-based methods are limited due to expensive sensors and time-consuming computation. Camera-based methods usually need to perform road segmentation and view transformation separately, which often causes distortion and missing content. To push the limits of the technology, we present a novel framework that reconstructs a local map formed by road layout and vehicle occupancy in the bird's-eye view given a front-view monocular image only. We propose a front-to-top view projection (FTVP) module, which takes the constraint of cycle consistency between views into account and makes full use of their correlation to strengthen the view transformation and scene understanding. In addition, we also apply multi-scale FTVP modules to propagate the rich spatial information of low-level features to mitigate spatial deviation of the predicted object location. Experiments on public benchmarks show that our method achieves the state-of-the-art performance in the tasks of road layout estimation, vehicle occupancy estimation, and multi-class semantic estimation. For multi-class semantic estimation, in particular, our model outperforms all competitors by a large margin. Furthermore, our model runs at 25 FPS on a single GPU, which is efficient and applicable for real-time panorama HD map reconstruction.
We investigate transductive zero-shot point cloud semantic segmentation in this paper, where unseen class labels are unavailable during training. Actually, the 3D geometric elements are essential cues to reason the 3D object type. If two categories share similar geometric primitives, they also have similar semantic representations. Based on this consideration, we propose a novel framework to learn the geometric primitives shared in seen and unseen categories' objects, where the learned geometric primitives are served for transferring knowledge from seen to unseen categories. Specifically, a group of learnable prototypes automatically encode geometric primitives via back-propagation. Then, the point visual representation is formulated as the similarity vector of its feature to the prototypes, which implies semantic cues for both seen and unseen categories. Besides, considering a 3D object composed of multiple geometric primitives, we formulate the semantic representation as a mixture-distributed embedding for the fine-grained match of visual representation. In the end, to effectively learn the geometric primitives and alleviate the misclassification issue, we propose a novel unknown-aware infoNCE loss to align the visual and semantic representation. As a result, guided by semantic representations, the network recognizes the novel object represented with geometric primitives. Extensive experiments show that our method significantly outperforms other state-of-the-art methods in the harmonic mean-intersection-over-union (hIoU), with the improvement of 17.8%, 30.4% and 9.2% on S3DIS, ScanNet and SemanticKITTI datasets, respectively. Codes will be released.
Predicting the future motion of road participants is crucial for autonomous driving but is extremely challenging due to staggering motion uncertainty. Recently, most motion forecasting methods resort to the goal-based strategy, i.e., predicting endpoints of motion trajectories as conditions to regress the entire trajectories, so that the search space of solution can be reduced. However, accurate goal coordinates are hard to predict and evaluate. In addition, the point representation of the destination limits the utilization of a rich road context, leading to inaccurate prediction results in many cases. Goal area, i.e., the possible destination area, rather than goal coordinate, could provide a more soft constraint for searching potential trajectories by involving more tolerance and guidance. In view of this, we propose a new goal area-based framework, named Goal Area Network (GANet), for motion forecasting, which models goal areas rather than exact goal coordinates as preconditions for trajectory prediction, performing more robustly and accurately. Specifically, we propose a GoICrop (Goal Area of Interest) operator to effectively extract semantic lane features in goal areas and model actors' future interactions, which benefits a lot for future trajectory estimations. GANet ranks the 1st on the leaderboard of Argoverse Challenge among all public literature (till the paper submission), and its source codes will be released.
Vision-centric BEV perception has recently received increased attention from both industry and academia due to its inherent merits, including presenting a natural representation of the world and being fusion-friendly. With the rapid development of deep learning, numerous methods have been proposed to address the vision-centric BEV perception. However, there is no recent survey for this novel and growing research field. To stimulate its future research, this paper presents a comprehensive survey of recent progress of vision-centric BEV perception and its extensions. It collects and organizes the recent knowledge, and gives a systematic review and summary of commonly used algorithms. It also provides in-depth analyses and comparative results on several BEV perception tasks, facilitating the comparisons of future works and inspiring future research directions. Moreover, empirical implementation details are also discussed and shown to benefit the development of related algorithms.
This article addresses the problem of distilling knowledge from a large teacher model to a slim student network for LiDAR semantic segmentation. Directly employing previous distillation approaches yields inferior results due to the intrinsic challenges of point cloud, i.e., sparsity, randomness and varying density. To tackle the aforementioned problems, we propose the Point-to-Voxel Knowledge Distillation (PVD), which transfers the hidden knowledge from both point level and voxel level. Specifically, we first leverage both the pointwise and voxelwise output distillation to complement the sparse supervision signals. Then, to better exploit the structural information, we divide the whole point cloud into several supervoxels and design a difficulty-aware sampling strategy to more frequently sample supervoxels containing less-frequent classes and faraway objects. On these supervoxels, we propose inter-point and inter-voxel affinity distillation, where the similarity information between points and voxels can help the student model better capture the structural information of the surrounding environment. We conduct extensive experiments on two popular LiDAR segmentation benchmarks, i.e., nuScenes and SemanticKITTI. On both benchmarks, our PVD consistently outperforms previous distillation approaches by a large margin on three representative backbones, i.e., Cylinder3D, SPVNAS and MinkowskiNet. Notably, on the challenging nuScenes and SemanticKITTI datasets, our method can achieve roughly 75% MACs reduction and 2x speedup on the competitive Cylinder3D model and rank 1st on the SemanticKITTI leaderboard among all published algorithms. Our code is available at https://github.com/cardwing/Codes-for-PVKD.