Zixiang Zhou

Zixiang Zhou, Xiangchen Zhao, Yu Wang et al.

Query-based transformer has shown great potential in constructing long-range attention in many image-domain tasks, but has rarely been considered in LiDAR-based 3D object detection due to the overwhelming size of the point cloud data. In this paper, we propose CenterFormer, a center-based transformer network for 3D object detection. CenterFormer first uses a center heatmap to select center candidates on top of a standard voxel-based point cloud encoder. It then uses the feature of the center candidate as the query embedding in the transformer. To further aggregate features from multiple frames, we design an approach to fuse features through cross-attention. Lastly, regression heads are added to predict the bounding box on the output center feature representation. Our design reduces the convergence difficulty and computational complexity of the transformer structure. The results show significant improvements over the strong baseline of anchor-free object detection networks. CenterFormer achieves state-of-the-art performance for a single model on the Waymo Open Dataset, with 73.7% mAPH on the validation set and 75.6% mAPH on the test set, significantly outperforming all previously published CNN and transformer-based methods. Our code is publicly available at https://github.com/TuSimple/centerformer

Zixiang Zhou, Baoyuan Wang

Generating controllable and editable human motion sequences is a key challenge in 3D Avatar generation. It has been labor-intensive to generate and animate human motion for a long time until learning-based approaches have been developed and applied recently. However, these approaches are still task-specific or modality-specific\cite {ahuja2019language2pose}\cite{ghosh2021synthesis}\cite{ferreira2021learning}\cite{li2021ai}. In this paper, we propose ``UDE", the first unified driving engine that enables generating human motion sequences from natural language or audio sequences (see Fig.~\ref{fig:teaser}). Specifically, UDE consists of the following key components: 1) a motion quantization module based on VQVAE that represents continuous motion sequence as discrete latent code\cite{van2017neural}, 2) a modality-agnostic transformer encoder\cite{vaswani2017attention} that learns to map modality-aware driving signals to a joint space, and 3) a unified token transformer (GPT-like\cite{radford2019language}) network to predict the quantized latent code index in an auto-regressive manner. 4) a diffusion motion decoder that takes as input the motion tokens and decodes them into motion sequences with high diversity. We evaluate our method on HumanML3D\cite{Guo_2022_CVPR} and AIST++\cite{li2021learn} benchmarks, and the experiment results demonstrate our method achieves state-of-the-art performance. Project website: \url{https://github.com/zixiangzhou916/UDE/

Shengnan Hu, Ce Zheng, Zixiang Zhou et al.

Human-centric visual understanding is an important desideratum for effective human-robot interaction. In order to navigate crowded public places, social robots must be able to interpret the activity of the surrounding humans. This paper addresses one key aspect of human-centric visual understanding, multi-person pose estimation. Achieving good performance on multi-person pose estimation in crowded scenes is difficult due to the challenges of occluded joints and instance separation. In order to tackle these challenges and overcome the limitations of image features in representing invisible body parts, we propose a novel prompt-based pose inference strategy called LAMP (Language Assisted Multi-person Pose estimation). By utilizing the text representations generated by a well-trained language model (CLIP), LAMP can facilitate the understanding of poses on the instance and joint levels, and learn more robust visual representations that are less susceptible to occlusion. This paper demonstrates that language-supervised training boosts the performance of single-stage multi-person pose estimation, and both instance-level and joint-level prompts are valuable for training. The code is available at https://github.com/shengnanh20/LAMP.

Zixiang Zhou, Dongqiangzi Ye, Weijia Chen et al.

There is a recent trend in the LiDAR perception field towards unifying multiple tasks in a single strong network with improved performance, as opposed to using separate networks for each task. In this paper, we introduce a new LiDAR multi-task learning paradigm based on the transformer. The proposed LiDARFormer utilizes cross-space global contextual feature information and exploits cross-task synergy to boost the performance of LiDAR perception tasks across multiple large-scale datasets and benchmarks. Our novel transformer-based framework includes a cross-space transformer module that learns attentive features between the 2D dense Bird's Eye View (BEV) and 3D sparse voxel feature maps. Additionally, we propose a transformer decoder for the segmentation task to dynamically adjust the learned features by leveraging the categorical feature representations. Furthermore, we combine the segmentation and detection features in a shared transformer decoder with cross-task attention layers to enhance and integrate the object-level and class-level features. LiDARFormer is evaluated on the large-scale nuScenes and the Waymo Open datasets for both 3D detection and semantic segmentation tasks, and it outperforms all previously published methods on both tasks. Notably, LiDARFormer achieves the state-of-the-art performance of 76.4% L2 mAPH and 74.3% NDS on the challenging Waymo and nuScenes detection benchmarks for a single model LiDAR-only method.

Dongqiangzi Ye, Yufei Xie, Weijia Chen et al.

Due to the difficulty of acquiring large-scale 3D human keypoint annotation, previous methods for 3D human pose estimation (HPE) have often relied on 2D image features and sequential 2D annotations. Furthermore, the training of these networks typically assumes the prediction of a human bounding box and the accurate alignment of 3D point clouds with 2D images, making direct application in real-world scenarios challenging. In this paper, we present the 1st framework for end-to-end 3D human pose estimation, named LPFormer, which uses only LiDAR as its input along with its corresponding 3D annotations. LPFormer consists of two stages: firstly, it identifies the human bounding box and extracts multi-level feature representations, and secondly, it utilizes a transformer-based network to predict human keypoints based on these features. Our method demonstrates that 3D HPE can be seamlessly integrated into a strong LiDAR perception network and benefit from the features extracted by the network. Experimental results on the Waymo Open Dataset demonstrate the state-of-the-art performance, and improvements even compared to previous multi-modal solutions.

Minkyoung Cho, Yulong Cao, Zixiang Zhou et al.

Deep neural networks (DNNs) are increasingly integrated into LiDAR (Light Detection and Ranging)-based perception systems for autonomous vehicles (AVs), requiring robust performance under adversarial conditions. We aim to address the challenge of LiDAR spoofing attacks, where attackers inject fake objects into LiDAR data and fool AVs to misinterpret their environment and make erroneous decisions. However, current defense algorithms predominantly depend on perception outputs (i.e., bounding boxes) thus face limitations in detecting attackers given the bounding boxes are generated by imperfect perception models processing limited points, acquired based on the ego vehicle's viewpoint. To overcome these limitations, we propose a novel framework, named ADoPT (Anomaly Detection based on Point-level Temporal consistency), which quantitatively measures temporal consistency across consecutive frames and identifies abnormal objects based on the coherency of point clusters. In our evaluation using the nuScenes dataset, our algorithm effectively counters various LiDAR spoofing attacks, achieving a low (< 10%) false positive ratio (FPR) and high (> 85%) true positive ratio (TPR), outperforming existing state-of-the-art defense methods, CARLO and 3D-TC2. Furthermore, our evaluation demonstrates the promising potential for accurate attack detection across various road environments.

Zixiang Zhou, Yu Wan, Baoyuan Wang

Large Language Models(LLMs) have shown remarkable emergent abilities in unifying almost all (if not every) NLP tasks. In the human motion-related realm, however, researchers still develop siloed models for each task. Inspired by InstuctGPT, and the generalist concept behind Gato, we introduce AvatarGPT, an All-in-One framework for motion understanding, planning, generations as well as other tasks such as motion in-between synthesis. AvatarGPT treats each task as one type of instruction fine-tuned on the shared LLM. All the tasks are seamlessly interconnected with language as the universal interface, constituting a closed-loop within the framework. To achieve this, human motion sequences are first encoded as discrete tokens, which serve as the extended vocabulary of LLM. Then, an unsupervised pipeline to generate natural language descriptions of human action sequences from in-the-wild videos is developed. Finally, all tasks are jointly trained. Extensive experiments show that AvatarGPT achieves SOTA on low-level tasks, and promising results on high-level tasks, demonstrating the effectiveness of our proposed All-in-One framework. Moreover, for the first time, AvatarGPT enables a principled approach by iterative traversal of the tasks within the closed-loop for unlimited long-motion synthesis.

Zixiang Zhou, Yu Wan, Baoyuan Wang

The field has made significant progress in synthesizing realistic human motion driven by various modalities. Yet, the need for different methods to animate various body parts according to different control signals limits the scalability of these techniques in practical scenarios. In this paper, we introduce a cohesive and scalable approach that consolidates multimodal (text, music, speech) and multi-part (hand, torso) human motion generation. Our methodology unfolds in several steps: We begin by quantizing the motions of diverse body parts into separate codebooks tailored to their respective domains. Next, we harness the robust capabilities of pre-trained models to transcode multimodal signals into a shared latent space. We then translate these signals into discrete motion tokens by iteratively predicting subsequent tokens to form a complete sequence. Finally, we reconstruct the continuous actual motion from this tokenized sequence. Our method frames the multimodal motion generation challenge as a token prediction task, drawing from specialized codebooks based on the modality of the control signal. This approach is inherently scalable, allowing for the easy integration of new modalities. Extensive experiments demonstrated the effectiveness of our design, emphasizing its potential for broad application.

Zixiang Zhou, Weiyuan Li, Baoyuan Wang

We propose MDSC(Music-Dance-Style Consistency), the first evaluation metric that assesses to what degree the dance moves and music match. Existing metrics can only evaluate the motion fidelity and diversity and the degree of rhythmic matching between music and dance. MDSC measures how stylistically correlated the generated dance motion sequences and the conditioning music sequences are. We found that directly measuring the embedding distance between motion and music is not an optimal solution. We instead tackle this through modeling it as a clustering problem. Specifically, 1) we pre-train a music encoder and a motion encoder, then 2) we learn to map and align the motion and music embedding in joint space by jointly minimizing the intra-cluster distance and maximizing the inter-cluster distance, and 3) for evaluation purposes, we encode the dance moves into embedding and measure the intra-cluster and inter-cluster distances, as well as the ratio between them. We evaluate our metric on the results of several music-conditioned motion generation methods, combined with user study, we found that our proposed metric is a robust evaluation metric in measuring the music-dance style correlation.

Zixiang Zhou, Yang Zhang, Hassan Foroosh

Panoptic segmentation presents a new challenge in exploiting the merits of both detection and segmentation, with the aim of unifying instance segmentation and semantic segmentation in a single framework. However, an efficient solution for panoptic segmentation in the emerging domain of LiDAR point cloud is still an open research problem and is very much under-explored. In this paper, we present a fast and robust LiDAR point cloud panoptic segmentation framework, referred to as Panoptic-PolarNet. We learn both semantic segmentation and class-agnostic instance clustering in a single inference network using a polar Bird's Eye View (BEV) representation, enabling us to circumvent the issue of occlusion among instances in urban street scenes. To improve our network's learnability, we also propose an adapted instance augmentation technique and a novel adversarial point cloud pruning method. Our experiments show that Panoptic-PolarNet outperforms the baseline methods on SemanticKITTI and nuScenes datasets with an almost real-time inference speed. Panoptic-PolarNet achieved 54.1% PQ in the public SemanticKITTI panoptic segmentation leaderboard and leading performance for the validation set of nuScenes.

Yang Zhang, Zixiang Zhou, Philip David et al.

The need for fine-grained perception in autonomous driving systems has resulted in recently increased research on online semantic segmentation of single-scan LiDAR. Despite the emerging datasets and technological advancements, it remains challenging due to three reasons: (1) the need for near-real-time latency with limited hardware; (2) uneven or even long-tailed distribution of LiDAR points across space; and (3) an increasing number of extremely fine-grained semantic classes. In an attempt to jointly tackle all the aforementioned challenges, we propose a new LiDAR-specific, nearest-neighbor-free segmentation algorithm - PolarNet. Instead of using common spherical or bird's-eye-view projection, our polar bird's-eye-view representation balances the points across grid cells in a polar coordinate system, indirectly aligning a segmentation network's attention with the long-tailed distribution of the points along the radial axis. We find that our encoding scheme greatly increases the mIoU in three drastically different segmentation datasets of real urban LiDAR single scans while retaining near real-time throughput.