Marcelo H. Ang

Ziyuan Huang, Shiwei Zhang, Liang Pan et al.

Spatial convolutions are extensively used in numerous deep video models. It fundamentally assumes spatio-temporal invariance, i.e., using shared weights for every location in different frames. This work presents Temporally-Adaptive Convolutions (TAdaConv) for video understanding, which shows that adaptive weight calibration along the temporal dimension is an efficient way to facilitate modeling complex temporal dynamics in videos. Specifically, TAdaConv empowers spatial convolutions with temporal modeling abilities by calibrating the convolution weights for each frame according to its local and global temporal context. Compared to existing operations for temporal modeling, TAdaConv is more efficient as it operates over the convolution kernels instead of the features, whose dimension is an order of magnitude smaller than the spatial resolutions. Further, kernel calibration brings an increased model capacity. Based on this readily plug-in operation TAdaConv as well as its extension, i.e., TAdaConvV2, we construct TAdaBlocks to empower ConvNeXt and Vision Transformer to have strong temporal modeling capabilities. Empirical results show TAdaConvNeXtV2 and TAdaFormer perform competitively against state-of-the-art convolutional and Transformer-based models in various video understanding benchmarks. Our codes and models are released at: https://github.com/alibaba-mmai-research/TAdaConv.

Yechao Bai, Xiaogang Wang, Marcelo H. Ang et al.

The learning and aggregation of multi-scale features are essential in empowering neural networks to capture the fine-grained geometric details in the point cloud upsampling task. Most existing approaches extract multi-scale features from a point cloud of a fixed resolution, hence obtain only a limited level of details. Though an existing approach aggregates a feature hierarchy of different resolutions from a cascade of upsampling sub-network, the training is complex with expensive computation. To address these issues, we construct a new point cloud upsampling pipeline called BIMS-PU that integrates the feature pyramid architecture with a bi-directional up and downsampling path. Specifically, we decompose the up/downsampling procedure into several up/downsampling sub-steps by breaking the target sampling factor into smaller factors. The multi-scale features are naturally produced in a parallel manner and aggregated using a fast feature fusion method. Supervision signal is simultaneously applied to all upsampled point clouds of different scales. Moreover, we formulate a residual block to ease the training of our model. Extensive quantitative and qualitative experiments on different datasets show that our method achieves superior results to state-of-the-art approaches. Last but not least, we demonstrate that point cloud upsampling can improve robot perception by ameliorating the 3D data quality.

Lei Zhou, Zhiyang Liu, Runze Gan et al.

Category-level object pose estimation involves estimating the 6D pose and the 3D metric size of objects from predetermined categories. While recent approaches take categorical shape prior information as reference to improve pose estimation accuracy, the single-stage network design and training manner lead to sub-optimal performance since there are two distinct tasks in the pipeline. In this paper, the advantage of two-stage pipeline over single-stage design is discussed. To this end, we propose a two-stage deformation-and registration pipeline called DR-Pose, which consists of completion-aided deformation stage and scaled registration stage. The first stage uses a point cloud completion method to generate unseen parts of target object, guiding subsequent deformation on the shape prior. In the second stage, a novel registration network is designed to extract pose-sensitive features and predict the representation of object partial point cloud in canonical space based on the deformation results from the first stage. DR-Pose produces superior results to the state-of-the-art shape prior-based methods on both CAMERA25 and REAL275 benchmarks. Codes are available at https://github.com/Zray26/DR-Pose.git.

Zhili Ng, Haozhe Wang, Zhengshen Zhang et al.

In this work, we present SynTable, a unified and flexible Python-based dataset generator built using NVIDIA's Isaac Sim Replicator Composer for generating high-quality synthetic datasets for unseen object amodal instance segmentation of cluttered tabletop scenes. Our dataset generation tool can render complex 3D scenes containing object meshes, materials, textures, lighting, and backgrounds. Metadata, such as modal and amodal instance segmentation masks, object amodal RGBA instances, occlusion masks, depth maps, bounding boxes, and material properties can be automatically generated to annotate the scene according to the users' requirements. Our tool eliminates the need for manual labeling in the dataset generation process while ensuring the quality and accuracy of the dataset. In this work, we discuss our design goals, framework architecture, and the performance of our tool. We demonstrate the use of a sample dataset generated using SynTable for training a state-of-the-art model, UOAIS-Net. Our state-of-the-art results show significantly improved performance in Sim-to-Real transfer when evaluated on the OSD-Amodal dataset. We offer this tool as an open-source, easy-to-use, photorealistic dataset generator for advancing research in deep learning and synthetic data generation. The links to our source code, demonstration video, and sample dataset can be found in the supplementary materials.

Dongen Li, Yi Liu, Junqi Liu et al.

Achieving real-time Simultaneous Localization and Mapping (SLAM) based on 3D Gaussian splatting (3DGS) in large-scale real-world environments remains challenging, as existing methods still struggle to jointly achieve low-latency pose estimation, continuous 3D Gaussian reconstruction, and long-term global consistency. In this paper, we present a tightly coupled LiDAR-Inertial-Visual 3DGS-based SLAM framework for real-time pose estimation and photorealistic mapping in large-scale real-world scenes. The system executes state estimation and 3D Gaussian primitive initialization in parallel with global Gaussian optimization, enabling continuous dense mapping. To improve Gaussian initialization quality and accelerate optimization convergence, we introduce a cascaded strategy that combines feed-forward predictions with geometric priors derived from voxel-based principal component analysis. To enhance global consistency, we perform loop closure directly on the optimized global Gaussian map by estimating loop constraints through Gaussian-based Generalized Iterative Closest Point registration, followed by pose-graph optimization. We also collect challenging large-scale looped outdoor sequences with hardware-synchronized LiDAR-camera-IMU and ground-truth trajectories for realistic evaluation. Extensive experiments on both public datasets and our dataset demonstrate that the proposed method achieves a state of the art among real-time efficiency, localization accuracy, and rendering quality across diverse real-world scenes.

Chialoon Cheng, Kaijun liu, Zhiyang Liu et al.

This comprehensive review examines the evolution and the current state of the art in three-dimensional (3D) reconstruction techniques in manufacturing applications. The analysis covers both traditional approaches and emerging deep learning methods, showing a critical research gap in unified 3d reconstruction frameworks. Through systematic review of 106 recent publications, we classify reconstruction techniques into three primary categories: data acquisition, point cloud generation, post-processing and applications. Non-contact methods, particularly structured light scanning and stereo vision, have shown significant adoption in manufacturing, with 47% of surveyed applications focusing on quality inspection. The integration of deep learning has enhanced reconstruction accuracy and processing speed, particularly in feature extraction and matching. Key applications span design and development (13%), machining (8%), process (17%), assembly (22%), and quality inspection (40%). While current technologies achieve sub-millimeter accuracy in controlled environments, challenges persist in handling reflective surfaces and dynamic environments. Our findings indicate a trend toward hybrid systems combining multiple sensor types and processing methods to overcome individual limitations. This survey provides a structured framework for understanding current capabilities and future directions in manufacturing-focused 3D reconstruction.

Jay Karhade, Haiyue Zhu, Ka-Shing Chung et al.

We present a neural point cloud rendering pipeline through a novel multi-frequency-aware patch adversarial learning framework. The proposed approach aims to improve the rendering realness by minimizing the spectrum discrepancy between real and synthesized images, especially on the high-frequency localized sharpness information which causes image blur visually. Specifically, a patch multi-discriminator scheme is proposed for the adversarial learning, which combines both spectral domain (Fourier Transform and Discrete Wavelet Transform) discriminators as well as the spatial (RGB) domain discriminator to force the generator to capture global and local spectral distributions of the real images. The proposed multi-discriminator scheme not only helps to improve rendering realness, but also enhance the convergence speed and stability of adversarial learning. Moreover, we introduce a noise-resistant voxelisation approach by utilizing both the appearance distance and spatial distance to exclude the spatial outlier points caused by depth noise. Our entire architecture is fully differentiable and can be learned in an end-to-end fashion. Extensive experiments show that our method produces state-of-the-art results for neural point cloud rendering by a significant margin. Our source code will be made public at a later date.

Shuo Sun, Zekai Gu, Tianchen Sun et al.

Realistic and diverse traffic scenarios in large quantities are crucial for the development and validation of autonomous driving systems. However, owing to numerous difficulties in the data collection process and the reliance on intensive annotations, real-world datasets lack sufficient quantity and diversity to support the increasing demand for data. This work introduces DriveSceneGen, a data-driven driving scenario generation method that learns from the real-world driving dataset and generates entire dynamic driving scenarios from scratch. DriveSceneGen is able to generate novel driving scenarios that align with real-world data distributions with high fidelity and diversity. Experimental results on 5k generated scenarios highlight the generation quality, diversity, and scalability compared to real-world datasets. To the best of our knowledge, DriveSceneGen is the first method that generates novel driving scenarios involving both static map elements and dynamic traffic participants from scratch.

Yiwen Chen, Xue Li, Sheng Guo et al.

Reinforcement learning has shown a wide usage in robotics tasks, such as insertion and grasping. However, without a practical sim2real strategy, the policy trained in simulation could fail on the real task. There are also wide researches in the sim2real strategies, but most of those methods rely on heavy image rendering, domain randomization training, or tuning. In this work, we solve the insertion task using a pure visual reinforcement learning solution with minimum infrastructure requirement. We also propose a novel sim2real strategy, Real2Sim, which provides a novel and easier solution in policy adaptation. We discuss the advantage of Real2Sim compared with Sim2Real.

Liu Haofeng, Chen Yiwen, Tan Jiayi et al.

Combined with demonstrations, deep reinforcement learning can efficiently develop policies for manipulators. However, it takes time to collect sufficient high-quality demonstrations in practice. And human demonstrations may be unsuitable for robots. The non-Markovian process and over-reliance on demonstrations are further challenges. For example, we found that RL agents are sensitive to demonstration quality in manipulation tasks and struggle to adapt to demonstrations directly from humans. Thus it is challenging to leverage low-quality and insufficient demonstrations to assist reinforcement learning in training better policies, and sometimes, limited demonstrations even lead to worse performance. We propose a new algorithm named TD3fG (TD3 learning from a generator) to solve these problems. It forms a smooth transition from learning from experts to learning from experience. This innovation can help agents extract prior knowledge while reducing the detrimental effects of the demonstrations. Our algorithm performs well in Adroit manipulator and MuJoCo tasks with limited demonstrations.

Yiwen Chen, Sheng Guo, Zedong Zhang et al.

Precision robotic manipulation tasks (insertion, screwing, precisely pick, precisely place) are required in many scenarios. Previous methods achieved good performance on such manipulation tasks. However, such methods typically require tedious calibration or expensive sensors. 3D/RGB-D cameras and torque/force sensors add to the cost of the robotic application and may not always be economical. In this work, we aim to solve these but using only weak-calibrated and low-cost webcams. We propose Binocular Alignment Learning (BAL), which could automatically learn the eye-hand coordination and points alignment capabilities to solve the four tasks. Our work focuses on working with unknown eye-hand coordination and proposes different ways of performing eye-in-hand camera calibration automatically. The algorithm was trained in simulation and used a practical pipeline to achieve sim2real and test it on the real robot. Our method achieves a competitively good result with minimal cost on the four tasks.

Yiwen Chen, Zedong Zhang, Haofeng Liu et al.

In recent years, a myriad of superlative works on intelligent robotics policies have been done, thanks to advances in machine learning. However, inefficiency and lack of transfer ability hindered algorithms from pragmatic applications, especially in human-robot collaboration, when few-shot fast learning and high flexibility become a wherewithal. To surmount this obstacle, we refer to a "Policy Pool", containing pre-trained skills that can be easily accessed and reused. An agent is employed to govern the "Policy Pool" by unfolding requisite skills in a flexible sequence, contingent on task specific predilection. This predilection can be automatically interpreted from one or few human expert demonstrations. Under this hierarchical setting, our algorithm is able to pick up a sparse-reward, multi-stage knack with only one demonstration in a Mini-Grid environment, showing the potential for instantly mastering complex robotics skills from human instructors. Additionally, the innate quality of our algorithm also allows for lifelong learning, making it a versatile agent.

Haozhe Wang, Zhiyang Liu, Lei Zhou et al.

Vision-based grasp estimation is an essential part of robotic manipulation tasks in the real world. Existing planar grasp estimation algorithms have been demonstrated to work well in relatively simple scenes. But when it comes to complex scenes, such as cluttered scenes with messy backgrounds and moving objects, the algorithms from previous works are prone to generate inaccurate and unstable grasping contact points. In this work, we first study the existing planar grasp estimation algorithms and analyze the related challenges in complex scenes. Secondly, we design a Pixel-wise Efficient Grasp Generation Network (PEGG-Net) to tackle the problem of grasping in complex scenes. PEGG-Net can achieve improved state-of-the-art performance on the Cornell dataset (98.9%) and second-best performance on the Jacquard dataset (93.8%), outperforming other existing algorithms without the introduction of complex structures. Thirdly, PEGG-Net could operate in a closed-loop manner for added robustness in dynamic environments using position-based visual servoing (PBVS). Finally, we conduct real-world experiments on static, dynamic, and cluttered objects in different complex scenes. The results show that our proposed network achieves a high success rate in grasping irregular objects, household objects, and workshop tools. To benefit the community, our trained model and supplementary materials are available at https://github.com/HZWang96/PEGG-Net.

Xiaogang Wang, Marcelo H Ang, Gim Hee Lee

Deep learning technique has yielded significant improvements in point cloud completion with the aim of completing missing object shapes from partial inputs. However, most existing methods fail to recover realistic structures due to over-smoothing of fine-grained details. In this paper, we develop a voxel-based network for point cloud completion by leveraging edge generation (VE-PCN). We first embed point clouds into regular voxel grids, and then generate complete objects with the help of the hallucinated shape edges. This decoupled architecture together with a multi-scale grid feature learning is able to generate more realistic on-surface details. We evaluate our model on the publicly available completion datasets and show that it outperforms existing state-of-the-art approaches quantitatively and qualitatively. Our source code is available at https://github.com/xiaogangw/VE-PCN.

Xiaogang Wang, Marcelo H Ang, Gim Hee Lee

In view of the difficulty in reconstructing object details in point cloud completion, we propose a shape prior learning method for object completion. The shape priors include geometric information in both complete and the partial point clouds. We design a feature alignment strategy to learn the shape prior from complete points, and a coarse to fine strategy to incorporate partial prior in the fine stage. To learn the complete objects prior, we first train a point cloud auto-encoder to extract the latent embeddings from complete points. Then we learn a mapping to transfer the point features from partial points to that of the complete points by optimizing feature alignment losses. The feature alignment losses consist of a L2 distance and an adversarial loss obtained by Maximum Mean Discrepancy Generative Adversarial Network (MMD-GAN). The L2 distance optimizes the partial features towards the complete ones in the feature space, and MMD-GAN decreases the statistical distance of two point features in a Reproducing Kernel Hilbert Space. We achieve state-of-the-art performances on the point cloud completion task. Our code is available at https://github.com/xiaogangw/point-cloud-completion-shape-prior.

Meng Tian, Marcelo H Ang, Gim Hee Lee

We present a novel learning approach to recover the 6D poses and sizes of unseen object instances from an RGB-D image. To handle the intra-class shape variation, we propose a deep network to reconstruct the 3D object model by explicitly modeling the deformation from a pre-learned categorical shape prior. Additionally, our network infers the dense correspondences between the depth observation of the object instance and the reconstructed 3D model to jointly estimate the 6D object pose and size. We design an autoencoder that trains on a collection of object models and compute the mean latent embedding for each category to learn the categorical shape priors. Extensive experiments on both synthetic and real-world datasets demonstrate that our approach significantly outperforms the state of the art. Our code is available at https://github.com/mentian/object-deformnet.

Feng Xue, Guirong Zhuo, Ziyuan Huang et al.

In recent years, self-supervised methods for monocular depth estimation has rapidly become an significant branch of depth estimation task, especially for autonomous driving applications. Despite the high overall precision achieved, current methods still suffer from a) imprecise object-level depth inference and b) uncertain scale factor. The former problem would cause texture copy or provide inaccurate object boundary, and the latter would require current methods to have an additional sensor like LiDAR to provide depth ground-truth or stereo camera as additional training inputs, which makes them difficult to implement. In this work, we propose to address these two problems together by introducing DNet. Our contributions are twofold: a) a novel dense connected prediction (DCP) layer is proposed to provide better object-level depth estimation and b) specifically for autonomous driving scenarios, dense geometrical constrains (DGC) is introduced so that precise scale factor can be recovered without additional cost for autonomous vehicles. Extensive experiments have been conducted and, both DCP layer and DGC module are proved to be effectively solving the aforementioned problems respectively. Thanks to DCP layer, object boundary can now be better distinguished in the depth map and the depth is more continues on object level. It is also demonstrated that the performance of using DGC to perform scale recovery is comparable to that using ground-truth information, when the camera height is given and the ground point takes up more than 1.03\% of the pixels. Code is available at https://github.com/TJ-IPLab/DNet.

Xiaogang Wang, Marcelo H Ang, Gim Hee Lee

Point clouds are often sparse and incomplete. Existing shape completion methods are incapable of generating details of objects or learning the complex point distributions. To this end, we propose a cascaded refinement network together with a coarse-to-fine strategy to synthesize the detailed object shapes. Considering the local details of partial input with the global shape information together, we can preserve the existing details in the incomplete point set and generate the missing parts with high fidelity. We also design a patch discriminator that guarantees every local area has the same pattern with the ground truth to learn the complicated point distribution. Quantitative and qualitative experiments on different datasets show that our method achieves superior results compared to existing state-of-the-art approaches on the 3D point cloud completion task. Our source code is available at https://github.com/xiaogangw/cascaded-point-completion.git.

Meng Tian, Liang Pan, Marcelo H Ang et al.

Accurate 6D object pose estimation is fundamental to robotic manipulation and grasping. Previous methods follow a local optimization approach which minimizes the distance between closest point pairs to handle the rotation ambiguity of symmetric objects. In this work, we propose a novel discrete-continuous formulation for rotation regression to resolve this local-optimum problem. We uniformly sample rotation anchors in SO(3), and predict a constrained deviation from each anchor to the target, as well as uncertainty scores for selecting the best prediction. Additionally, the object location is detected by aggregating point-wise vectors pointing to the 3D center. Experiments on two benchmarks: LINEMOD and YCB-Video, show that the proposed method outperforms state-of-the-art approaches. Our code is available at https://github.com/mentian/object-posenet.

Jiahui Li, Tianle Shen, Zekai Gu et al.

Motion prediction is a challenging problem in autonomous driving as it demands the system to comprehend stochastic dynamics and the multi-modal nature of real-world agent interactions. Diffusion models have recently risen to prominence, and have proven particularly effective in pedestrian motion prediction tasks. However, the significant time consumption and sensitivity to noise have limited the real-time predictive capability of diffusion models. In response to these impediments, we propose a novel diffusion-based, acceleratable framework that adeptly predicts future trajectories of agents with enhanced resistance to noise. The core idea of our model is to learn a coarse-grained prior distribution of trajectory, which can skip a large number of denoise steps. This advancement not only boosts sampling efficiency but also maintains the fidelity of prediction accuracy. Our method meets the rigorous real-time operational standards essential for autonomous vehicles, enabling prompt trajectory generation that is vital for secure and efficient navigation. Through extensive experiments, our method speeds up the inference time to 136ms compared to standard diffusion model, and achieves significant improvement in multi-agent motion prediction on the Argoverse 1 motion forecasting dataset.

Peiyuan Ni, Chee Meng Chew, Marcelo H. Ang et al.

Bin-picking of metal objects using low-cost RGB-D cameras often suffers from sparse depth information and reflective surface textures, leading to errors and the need for manual labeling. To reduce human intervention, we propose a two-stage framework consisting of a metric learning stage and a self-training stage. Specifically, to automatically process data captured by a low-cost camera (LC), we introduce a Multi-object Pose Reasoning (MoPR) algorithm that optimizes pose hypotheses under depth, collision, and boundary constraints. To further refine pose candidates, we adopt a Symmetry-aware Lie-group based Bayesian Gaussian Mixture Model (SaL-BGMM), integrated with the Expectation-Maximization (EM) algorithm, for symmetry-aware filtering. Additionally, we propose a Weighted Ranking Information Noise Contrastive Estimation (WR-InfoNCE) loss to enable the LC to learn a perceptual metric from reconstructed data, supporting self-training on untrained or even unseen objects. Experimental results show that our approach outperforms several state-of-the-art methods on both the ROBI dataset and our newly introduced Self-ROBI dataset.

Xunjie He, Christina Dao Wen Lee, Meiling Wang et al.

Collaborative perception plays a crucial role in enhancing environmental understanding by expanding the perceptual range and improving robustness against sensor failures, which primarily involves collaborative 3D detection and tracking tasks. The former focuses on object recognition in individual frames, while the latter captures continuous instance tracklets over time. However, existing works in both areas predominantly focus on the vehicle superclass, lacking effective solutions for both multi-class collaborative detection and tracking. This limitation hinders their applicability in real-world scenarios, which involve diverse object classes with varying appearances and motion patterns. To overcome these limitations, we propose a multi-class collaborative detection and tracking framework tailored for diverse road users. We first present a detector with a global spatial attention fusion (GSAF) module, enhancing multi-scale feature learning for objects of varying sizes. Next, we introduce a tracklet RE-IDentification (REID) module that leverages visual semantics with a vision foundation model to effectively reduce ID SWitch (IDSW) errors, in cases of erroneous mismatches involving small objects like pedestrians. We further design a velocity-based adaptive tracklet management (VATM) module that adjusts the tracking interval dynamically based on object motion. Extensive experiments on the V2X-Real and OPV2V datasets show that our approach significantly outperforms existing state-of-the-art methods in both detection and tracking accuracy.

Shuo Sun, Yimin Zhao, Christina Dao Wen Lee et al.

As the field progresses toward Artificial General Intelligence (AGI), there is a pressing need for more comprehensive and insightful evaluation frameworks that go beyond aggregate performance metrics. This paper introduces a unified rating system that jointly models the difficulty of individual test cases and the competency of AI models (or humans) across vision, language, and action domains. Unlike existing metrics that focus solely on models, our approach allows for fine-grained, difficulty-aware evaluations through competitive interactions between models and tasks, capturing both the long-tail distribution of real-world challenges and the competency gap between current models and full task mastery. We validate the generalizability and robustness of our system through extensive experiments on multiple established datasets and models across distinct AGI domains. The resulting rating distributions offer novel perspectives and interpretable insights into task difficulty, model progression, and the outstanding challenges that remain on the path to achieving full AGI task mastery.

Yiwen Chen, Yuyao Ye, Ziyi Chen et al.

Robotics learning highly relies on human expertise and efforts, such as demonstrations, design of reward functions in reinforcement learning, performance evaluation using human feedback, etc. However, reliance on human assistance can lead to expensive learning costs and make skill learning difficult to scale. In this work, we introduce the Large Language Model Supervised Robotics Text2Skill Autonomous Learning (ARO) framework, which aims to replace human participation in the robot skill learning process with large-scale language models that incorporate reward function design and performance evaluation. We provide evidence that our approach enables fully autonomous robot skill learning, capable of completing partial tasks without human intervention. Furthermore, we also analyze the limitations of this approach in task understanding and optimization stability.

Arijit Dasgupta, Jiafei Duan, Marcelo H. Ang et al.

Recent work in computer vision and cognitive reasoning has given rise to an increasing adoption of the Violation-of-Expectation (VoE) paradigm in synthetic datasets. Inspired by infant psychology, researchers are now evaluating a model's ability to label scenes as either expected or surprising with knowledge of only expected scenes. However, existing VoE-based 3D datasets in physical reasoning provide mainly vision data with little to no heuristics or inductive biases. Cognitive models of physical reasoning reveal infants create high-level abstract representations of objects and interactions. Capitalizing on this knowledge, we established a benchmark to study physical reasoning by curating a novel large-scale synthetic 3D VoE dataset armed with ground-truth heuristic labels of causally relevant features and rules. To validate our dataset in five event categories of physical reasoning, we benchmarked and analyzed human performance. We also proposed the Object File Physical Reasoning Network (OFPR-Net) which exploits the dataset's novel heuristics to outperform our baseline and ablation models. The OFPR-Net is also flexible in learning an alternate physical reality, showcasing its ability to learn universal causal relationships in physical reasoning to create systems with better interpretability.

Haofeng Liu, Yiwen Chen, Jiayi Tan et al.

How to make imitation learning more general when demonstrations are relatively limited has been a persistent problem in reinforcement learning (RL). Poor demonstrations lead to narrow and biased date distribution, non-Markovian human expert demonstration makes it difficult for the agent to learn, and over-reliance on sub-optimal trajectories can make it hard for the agent to improve its performance. To solve these problems we propose a new algorithm named TD3fG that can smoothly transition from learning from experts to learning from experience. Our algorithm achieves good performance in the MUJOCO environment with limited and sub-optimal demonstrations. We use behavior cloning to train the network as a reference action generator and utilize it in terms of both loss function and exploration noise. This innovation can help agents extract a priori knowledge from demonstrations while reducing the detrimental effects of the poor Markovian properties of the demonstrations. It has a better performance compared to the BC+ fine-tuning and DDPGfD approach, especially when the demonstrations are relatively limited. We call our method TD3fG meaning TD3 from a generator.

Ziyuan Huang, Shiwei Zhang, Liang Pan et al.

Spatial convolutions are widely used in numerous deep video models. It fundamentally assumes spatio-temporal invariance, i.e., using shared weights for every location in different frames. This work presents Temporally-Adaptive Convolutions (TAdaConv) for video understanding, which shows that adaptive weight calibration along the temporal dimension is an efficient way to facilitate modelling complex temporal dynamics in videos. Specifically, TAdaConv empowers the spatial convolutions with temporal modelling abilities by calibrating the convolution weights for each frame according to its local and global temporal context. Compared to previous temporal modelling operations, TAdaConv is more efficient as it operates over the convolution kernels instead of the features, whose dimension is an order of magnitude smaller than the spatial resolutions. Further, the kernel calibration brings an increased model capacity. We construct TAda2D and TAdaConvNeXt networks by replacing the 2D convolutions in ResNet and ConvNeXt with TAdaConv, which leads to at least on par or better performance compared to state-of-the-art approaches on multiple video action recognition and localization benchmarks. We also demonstrate that as a readily plug-in operation with negligible computation overhead, TAdaConv can effectively improve many existing video models with a convincing margin.

Arijit Dasgupta, Jiafei Duan, Marcelo H. Ang et al.

Recent work in cognitive reasoning and computer vision has engendered an increasing popularity for the Violation-of-Expectation (VoE) paradigm in synthetic datasets. Inspired by work in infant psychology, researchers have started evaluating a model's ability to discriminate between expected and surprising scenes as a sign of its reasoning ability. Existing VoE-based 3D datasets in physical reasoning only provide vision data. However, current cognitive models of physical reasoning by psychologists reveal infants create high-level abstract representations of objects and interactions. Capitalizing on this knowledge, we propose AVoE: a synthetic 3D VoE-based dataset that presents stimuli from multiple novel sub-categories for five event categories of physical reasoning. Compared to existing work, AVoE is armed with ground-truth labels of abstract features and rules augmented to vision data, paving the way for high-level symbolic predictions in physical reasoning tasks.

Zhiwu Qing, Ziyuan Huang, Shiwei Zhang et al.

The central idea of contrastive learning is to discriminate between different instances and force different views from the same instance to share the same representation. To avoid trivial solutions, augmentation plays an important role in generating different views, among which random cropping is shown to be effective for the model to learn a generalized and robust representation. Commonly used random crop operation keeps the distribution of the difference between two views unchanged along the training process. In this work, we show that adaptively controlling the disparity between two augmented views along the training process enhances the quality of the learned representation. Specifically, we present a parametric cubic cropping operation, ParamCrop, for video contrastive learning, which automatically crops a 3D cubic by differentiable 3D affine transformations. ParamCrop is trained simultaneously with the video backbone using an adversarial objective and learns an optimal cropping strategy from the data. The visualizations show that ParamCrop adaptively controls the center distance and the IoU between two augmented views, and the learned change in the disparity along the training process is beneficial to learning a strong representation. Extensive ablation studies demonstrate the effectiveness of the proposed ParamCrop on multiple contrastive learning frameworks and video backbones. Codes and models will be available.

Zhiwu Qing, Ziyuan Huang, Xiang Wang et al.

This technical report analyzes an egocentric video action detection method we used in the 2021 EPIC-KITCHENS-100 competition hosted in CVPR2021 Workshop. The goal of our task is to locate the start time and the end time of the action in the long untrimmed video, and predict action category. We adopt sliding window strategy to generate proposals, which can better adapt to short-duration actions. In addition, we show that classification and proposals are conflict in the same network. The separation of the two tasks boost the detection performance with high efficiency. By simply employing these strategy, we achieved 16.10\% performance on the test set of EPIC-KITCHENS-100 Action Detection challenge using a single model, surpassing the baseline method by 11.7\% in terms of average mAP.

Ziyuan Huang, Zhiwu Qing, Xiang Wang et al.

With the recent surge in the research of vision transformers, they have demonstrated remarkable potential for various challenging computer vision applications, such as image recognition, point cloud classification as well as video understanding. In this paper, we present empirical results for training a stronger video vision transformer on the EPIC-KITCHENS-100 Action Recognition dataset. Specifically, we explore training techniques for video vision transformers, such as augmentations, resolutions as well as initialization, etc. With our training recipe, a single ViViT model achieves the performance of 47.4\% on the validation set of EPIC-KITCHENS-100 dataset, outperforming what is reported in the original paper by 3.4%. We found that video transformers are especially good at predicting the noun in the verb-noun action prediction task. This makes the overall action prediction accuracy of video transformers notably higher than convolutional ones. Surprisingly, even the best video transformers underperform the convolutional networks on the verb prediction. Therefore, we combine the video vision transformers and some of the convolutional video networks and present our solution to the EPIC-KITCHENS-100 Action Recognition competition.

Yechao Bai, Ziyuan Huang, Lyuyu Shen et al.

Exploiting multi-scale features has shown great potential in tackling semantic segmentation problems. The aggregation is commonly done with sum or concatenation (concat) followed by convolutional (conv) layers. However, it fully passes down the high-level context to the following hierarchy without considering their interrelation. In this work, we aim to enable the low-level feature to aggregate the complementary context from adjacent high-level feature maps by a cross-scale pixel-to-region relation operation. We leverage cross-scale context propagation to make the long-range dependency capturable even by the high-resolution low-level features. To this end, we employ an efficient feature pyramid network to obtain multi-scale features. We propose a Relational Semantics Extractor (RSE) and Relational Semantics Propagator (RSP) for context extraction and propagation respectively. Then we stack several RSP into an RSP head to achieve the progressive top-down distribution of the context. Experiment results on two challenging datasets Cityscapes and COCO demonstrate that the RSP head performs competitively on both semantic segmentation and panoptic segmentation with high efficiency. It outperforms DeeplabV3 [1] by 0.7% with 75% fewer FLOPs (multiply-adds) in the semantic segmentation task.

Ziyuan Huang, Shiwei Zhang, Jianwen Jiang et al.

Motions are reflected in videos as the movement of pixels, and actions are essentially patterns of inconsistent motions between the foreground and the background. To well distinguish the actions, especially those with complicated spatio-temporal interactions, correctly locating the prominent motion areas is of crucial importance. However, most motion information in existing videos are difficult to label and training a model with good motion representations with supervision will thus require a large amount of human labour for annotation. In this paper, we address this problem by self-supervised learning. Specifically, we propose to learn Motion from Static Images (MoSI). The model learns to encode motion information by classifying pseudo motions generated by MoSI. We furthermore introduce a static mask in pseudo motions to create local motion patterns, which forces the model to additionally locate notable motion areas for the correct classification.We demonstrate that MoSI can discover regions with large motion even without fine-tuning on the downstream datasets. As a result, the learned motion representations boost the performance of tasks requiring understanding of complex scenes and motions, i.e., action recognition. Extensive experiments show the consistent and transferable improvements achieved by MoSI. Codes will be soon released.

Xiaogang Wang, Marcelo H Ang, Gim Hee Lee

Point clouds are often sparse and incomplete, which imposes difficulties for real-world applications. Existing shape completion methods tend to generate rough shapes without fine-grained details. Considering this, we introduce a two-branch network for shape completion. The first branch is a cascaded shape completion sub-network to synthesize complete objects, where we propose to use the partial input together with the coarse output to preserve the object details during the dense point reconstruction. The second branch is an auto-encoder to reconstruct the original partial input. The two branches share a same feature extractor to learn an accurate global feature for shape completion. Furthermore, we propose two strategies to enable the training of our network when ground truth data are not available. This is to mitigate the dependence of existing approaches on large amounts of ground truth training data that are often difficult to obtain in real-world applications. Additionally, our proposed strategies are also able to improve the reconstruction quality for fully supervised learning. We verify our approach in self-supervised, semi-supervised and fully supervised settings with superior performances. Quantitative and qualitative results on different datasets demonstrate that our method achieves more realistic outputs than state-of-the-art approaches on the point cloud completion task.

Zehui Meng, Qi Heng Ho, Zefan Huang et al.

Target detection and tracking provides crucial information for motion planning and decision making in autonomous driving. This paper proposes an online multi-object tracking (MOT) framework with tracking-by-detection for maneuvering vehicles under motion uncertainty in dynamic road context. We employ a point cloud based vehicle detector to provide real-time 3D bounding boxes of detected vehicles and conduct the online bipartite optimization of the maneuver-orientated data association between the detections and the targets. Kalman Filter (KF) is adopted as the backbone for multi-object tracking. In order to entertain the maneuvering uncertainty, we leverage the interacting multiple model (IMM) approach to obtain the \textit{a-posterior} residual as the cost for each association hypothesis, which is calculated with the hybrid model posterior (after mode-switch). Road context is integrated to conduct adjustments of the time varying transition probability matrix (TPM) of the IMM to regulate the maneuvers according to road segments and traffic sign/signals, with which the data association is performed in a unified spatial-temporal fashion. Experiments show our framework is able to effectively track multiple vehicles with maneuvers subject to dynamic road context and localization drift.

Mengdan Feng, Sixing Hu, Marcelo Ang et al.

Large-scale point cloud generated from 3D sensors is more accurate than its image-based counterpart. However, it is seldom used in visual pose estimation due to the difficulty in obtaining 2D-3D image to point cloud correspondences. In this paper, we propose the 2D3D-MatchNet - an end-to-end deep network architecture to jointly learn the descriptors for 2D and 3D keypoint from image and point cloud, respectively. As a result, we are able to directly match and establish 2D-3D correspondences from the query image and 3D point cloud reference map for visual pose estimation. We create our Oxford 2D-3D Patches dataset from the Oxford Robotcar dataset with the ground truth camera poses and 2D-3D image to point cloud correspondences for training and testing the deep network. Experimental results verify the feasibility of our approach.

Xinxin Du, Marcelo H. Ang, Sertac Karaman et al.

Autonomous driving requires 3D perception of vehicles and other objects in the in environment. Much of the current methods support 2D vehicle detection. This paper proposes a flexible pipeline to adopt any 2D detection network and fuse it with a 3D point cloud to generate 3D information with minimum changes of the 2D detection networks. To identify the 3D box, an effective model fitting algorithm is developed based on generalised car models and score maps. A two-stage convolutional neural network (CNN) is proposed to refine the detected 3D box. This pipeline is tested on the KITTI dataset using two different 2D detection networks. The 3D detection results based on these two networks are similar, demonstrating the flexibility of the proposed pipeline. The results rank second among the 3D detection algorithms, indicating its competencies in 3D detection.

Xiaotong Shen, Hans Andersen, Wei Kang Leong et al.

When a vehicle observes another one, the two vehicles' poses are correlated by this spatial relative observation, which can be used in cooperative localization for further increasing localization accuracy and precision. To use spatial relative observations, we propose to add them into a pose graph for optimal pose estimation. Before adding them, we need to know the identities of the observed vehicles. The vehicle identification is formulated as a linear assignment problem, which can be solved efficiently. By using pose graph techniques and the start-of-the-art factor composition/decomposition method, our cooperative localization algorithm is robust against communication delay, packet loss, and out-of-sequence packet reception. We demonstrate the usability of our framework and effectiveness of our algorithm through both simulations and real-world experiments using three vehicles on the road.