Pan Ji

Yang Liu, Jiyao Yang, Madhawa Perera et al.

3D skeletal action recognition has emerged as a powerful alternative to traditional RGB and depth-based approaches, offering robustness to environmental variations, computational efficiency, and enhanced privacy. Despite remarkable progress, current research remains fragmented across diverse input representations and lacks evaluation under scenarios that reflect real-world challenges. This paper presents a representation-centric review of supervised skeletal action recognition, systematically categorizing state-of-the-art methods by their input feature types: joint coordinates, bone vectors, motion flows, and extended representations, and analyzing how these choices influence spatiotemporal modeling strategies. Building on the insights from this review, we introduce ANUBIS, a large-scale, challenging dataset designed to address critical gaps in existing benchmarks. ANUBIS incorporates multi-view recordings with back-view perspectives, complex multi-person interactions, fine-grained and violent actions, and contemporary social behaviors. We benchmark a diverse set of state-of-the-art models on ANUBIS and conduct an in-depth analysis of how different feature types affect recognition performance across 102 action categories. Our results show strong action-feature dependencies, highlight the limitations of naive multi-representational fusion, and point toward the need for task-aware, semantically aligned integration strategies. This work offers both a comprehensive foundation and a practical benchmarking resource, aiming to guide the next generation of robust, generalizable skeleton-based action recognition systems for complex real-world scenarios. The dataset, benchmarking framework, and code are available at https://yliu1082.github.io/ANUBIS/.

Jiachen Liu, Pan Ji, Nitin Bansal et al.

We present a novel framework named PlaneMVS for 3D plane reconstruction from multiple input views with known camera poses. Most previous learning-based plane reconstruction methods reconstruct 3D planes from single images, which highly rely on single-view regression and suffer from depth scale ambiguity. In contrast, we reconstruct 3D planes with a multi-view-stereo (MVS) pipeline that takes advantage of multi-view geometry. We decouple plane reconstruction into a semantic plane detection branch and a plane MVS branch. The semantic plane detection branch is based on a single-view plane detection framework but with differences. The plane MVS branch adopts a set of slanted plane hypotheses to replace conventional depth hypotheses to perform plane sweeping strategy and finally learns pixel-level plane parameters and its planar depth map. We present how the two branches are learned in a balanced way, and propose a soft-pooling loss to associate the outputs of the two branches and make them benefit from each other. Extensive experiments on various indoor datasets show that PlaneMVS significantly outperforms state-of-the-art (SOTA) single-view plane reconstruction methods on both plane detection and 3D geometry metrics. Our method even outperforms a set of SOTA learning-based MVS methods thanks to the learned plane priors. To the best of our knowledge, this is the first work on 3D plane reconstruction within an end-to-end MVS framework. Source code: https://github.com/oppo-us-research/PlaneMVS.

Sudhir Yarram, Jialian Wu, Pan Ji et al.

Video instance segmentation (VIS) task requires classifying, segmenting, and tracking object instances over all frames in a video clip. Recently, VisTR has been proposed as end-to-end transformer-based VIS framework, while demonstrating state-of-the-art performance. However, VisTR is slow to converge during training, requiring around 1000 GPU hours due to the high computational cost of its transformer attention module. To improve the training efficiency, we propose Deformable VisTR, leveraging spatio-temporal deformable attention module that only attends to a small fixed set of key spatio-temporal sampling points around a reference point. This enables Deformable VisTR to achieve linear computation in the size of spatio-temporal feature maps. Moreover, it can achieve on par performance as the original VisTR with 10$\times$ less GPU training hours. We validate the effectiveness of our method on the Youtube-VIS benchmark. Code is available at https://github.com/skrya/DefVIS.

Runze Li, Pan Ji, Yi Xu et al.

Self-supervised monocular depth estimation has seen significant progress in recent years, especially in outdoor environments. However, depth prediction results are not satisfying in indoor scenes where most of the existing data are captured with hand-held devices. As compared to outdoor environments, estimating depth of monocular videos for indoor environments, using self-supervised methods, results in two additional challenges: (i) the depth range of indoor video sequences varies a lot across different frames, making it difficult for the depth network to induce consistent depth cues for training; (ii) the indoor sequences recorded with handheld devices often contain much more rotational motions, which cause difficulties for the pose network to predict accurate relative camera poses. In this work, we propose a novel framework-MonoIndoor++ by giving special considerations to those challenges and consolidating a set of good practices for improving the performance of self-supervised monocular depth estimation for indoor environments. First, a depth factorization module with transformer-based scale regression network is proposed to estimate a global depth scale factor explicitly, and the predicted scale factor can indicate the maximum depth values. Second, rather than using a single-stage pose estimation strategy as in previous methods, we propose to utilize a residual pose estimation module to estimate relative camera poses across consecutive frames iteratively. Third, to incorporate extensive coordinates guidance for our residual pose estimation module, we propose to perform coordinate convolutional encoding directly over the inputs to pose networks. The proposed method is validated on a variety of benchmark indoor datasets, i.e., EuRoC MAV, NYUv2, ScanNet and 7-Scenes, demonstrating the state-of-the-art performance.

Changjiang Cai, Pan Ji, Qingan Yan et al.

This paper presents a learning-based method for multi-view depth estimation from posed images. Our core idea is a "learning-to-optimize" paradigm that iteratively indexes a plane-sweeping cost volume and regresses the depth map via a convolutional Gated Recurrent Unit (GRU). Since the cost volume plays a paramount role in encoding the multi-view geometry, we aim to improve its construction both at pixel- and frame- levels. At the pixel level, we propose to break the symmetry of the Siamese network (which is typically used in MVS to extract image features) by introducing a transformer block to the reference image (but not to the source images). Such an asymmetric volume allows the network to extract global features from the reference image to predict its depth map. Given potential inaccuracies in the poses between reference and source images, we propose to incorporate a residual pose network to correct the relative poses. This essentially rectifies the cost volume at the frame level. We conduct extensive experiments on real-world MVS datasets and show that our method achieves state-of-the-art performance in terms of both within-dataset evaluation and cross-dataset generalization. Code available: https://github.com/oppo-us-research/riav-mvs.

Yang Liu, Jiyao Yang, Madhawa Perera et al.

3D skeleton-based human action recognition has emerged as a powerful alternative to traditional RGB and depth-based approaches, offering robustness to environmental variations, computational efficiency, and enhanced privacy. Despite remarkable progress, current research remains fragmented across diverse input representations and lacks evaluation under scenarios that reflect modern real-world challenges. This paper presents a representation-centric survey of skeleton-based action recognition, systematically categorizing state-of-the-art methods by their input feature types: joint coordinates, bone vectors, motion flows, and extended representations, and analyzing how these choices influence spatial-temporal modeling strategies. Building on the insights from this review, we introduce ANUBIS, a large-scale, challenging skeleton action dataset designed to address critical gaps in existing benchmarks. ANUBIS incorporates multi-view recordings with back-view perspectives, complex multi-person interactions, fine-grained and violent actions, and contemporary social behaviors. We benchmark a diverse set of state-of-the-art models on ANUBIS and conduct an in-depth analysis of how different feature types affect recognition performance across 102 action categories. Our results show strong action-feature dependencies, highlight the limitations of naïve multi-representational fusion, and point toward the need for task-aware, semantically aligned integration strategies. This work offers both a comprehensive foundation and a practical benchmarking resource, aiming to guide the next generation of robust, generalizable skeleton-based action recognition systems for complex real-world scenarios. The dataset website, benchmarking framework, and download link are available at https://yliu1082.github.io/ANUBIS/.

Nitin Bansal, Pan Ji, Junsong Yuan et al.

Multi-task learning (MTL) paradigm focuses on jointly learning two or more tasks, aiming for significant improvement w.r.t model's generalizability, performance, and training/inference memory footprint. The aforementioned benefits become ever so indispensable in the case of joint training for vision-related {\bf dense} prediction tasks. In this work, we tackle the MTL problem of two dense tasks, i.e., semantic segmentation and depth estimation, and present a novel attention module called Cross-Channel Attention Module ({CCAM}), which facilitates effective feature sharing along each channel between the two tasks, leading to mutual performance gain with a negligible increase in trainable parameters. In a true symbiotic spirit, we then formulate a novel data augmentation for the semantic segmentation task using predicted depth called {AffineMix}, and a simple depth augmentation using predicted semantics called {ColorAug}. Finally, we validate the performance gain of the proposed method on the Cityscapes and ScanNet dataset, which helps us achieve state-of-the-art results for a semi-supervised joint model based on depth and semantic segmentation.

Zhenyue Qin, Pan Ji, Dongwoo Kim et al.

Skeleton sequences are compact and lightweight. Numerous skeleton-based action recognizers have been proposed to classify human behaviors. In this work, we aim to incorporate components that are compatible with existing models and further improve their accuracy. To this end, we design two temporal accessories: discrete cosine encoding (DCE) and chronological loss (CRL). DCE facilitates models to analyze motion patterns from the frequency domain and meanwhile alleviates the influence of signal noise. CRL guides networks to explicitly capture the sequence's chronological order. These two components consistently endow many recently-proposed action recognizers with accuracy boosts, achieving new state-of-the-art (SOTA) accuracy on two large datasets.

Xiangyu Xu, Lichang Chen, Changjiang Cai et al.

Direct optimization of interpolated features on multi-resolution voxel grids has emerged as a more efficient alternative to MLP-like modules. However, this approach is constrained by higher memory expenses and limited representation capabilities. In this paper, we introduce a novel dynamic grid optimization method for high-fidelity 3D surface reconstruction that incorporates both RGB and depth observations. Rather than treating each voxel equally, we optimize the process by dynamically modifying the grid and assigning more finer-scale voxels to regions with higher complexity, allowing us to capture more intricate details. Furthermore, we develop a scheme to quantify the dynamic subdivision of voxel grid during optimization without requiring any priors. The proposed approach is able to generate high-quality 3D reconstructions with fine details on both synthetic and real-world data, while maintaining computational efficiency, which is substantially faster than the baseline method NeuralRGBD.

Pan Ji, Qingan Yan, Yuxin Ma et al.

We present a robust and accurate depth refinement system, named GeoRefine, for geometrically-consistent dense mapping from monocular sequences. GeoRefine consists of three modules: a hybrid SLAM module using learning-based priors, an online depth refinement module leveraging self-supervision, and a global mapping module via TSDF fusion. The proposed system is online by design and achieves great robustness and accuracy via: (i) a robustified hybrid SLAM that incorporates learning-based optical flow and/or depth; (ii) self-supervised losses that leverage SLAM outputs and enforce long-term geometric consistency; (iii) careful system design that avoids degenerate cases in online depth refinement. We extensively evaluate GeoRefine on multiple public datasets and reach as low as $5\%$ absolute relative depth errors.

Zhiqi Zhang, Nitin Bansal, Changjiang Cai et al.

Synthetic datasets are often used to pretrain end-to-end optical flow networks, due to the lack of a large amount of labeled, real-scene data. But major drops in accuracy occur when moving from synthetic to real scenes. How do we better transfer the knowledge learned from synthetic to real domains? To this end, we propose CLIP-FLow, a semi-supervised iterative pseudo-labeling framework to transfer the pretraining knowledge to the target real domain. We leverage large-scale, unlabeled real data to facilitate transfer learning with the supervision of iteratively updated pseudo-ground truth labels, bridging the domain gap between the synthetic and the real. In addition, we propose a contrastive flow loss on reference features and the warped features by pseudo ground truth flows, to further boost the accurate matching and dampen the mismatching due to motion, occlusion, or noisy pseudo labels. We adopt RAFT as the backbone and obtain an F1-all error of 4.11%, i.e. a 19% error reduction from RAFT (5.10%) and ranking 2$^{nd}$ place at submission on the KITTI 2015 benchmark. Our framework can also be extended to other models, e.g. CRAFT, reducing the F1-all error from 4.79% to 4.66% on KITTI 2015 benchmark.

Qingan Yan, Pan Ji, Nitin Bansal et al.

In this paper, we deal with the problem of monocular depth estimation for fisheye cameras in a self-supervised manner. A known issue of self-supervised depth estimation is that it suffers in low-light/over-exposure conditions and in large homogeneous regions. To tackle this issue, we propose a novel ordinal distillation loss that distills the ordinal information from a large teacher model. Such a teacher model, since having been trained on a large amount of diverse data, can capture the depth ordering information well, but lacks in preserving accurate scene geometry. Combined with self-supervised losses, we show that our model can not only generate reasonable depth maps in challenging environments but also better recover the scene geometry. We further leverage the fisheye cameras of an AR-Glasses device to collect an indoor dataset to facilitate evaluation.

Chengzeng Feng, Jiacheng Wei, Cheng Chen et al.

We propose Prim2Room, a novel framework for controllable room mesh generation leveraging 2D layout conditions and 3D primitive retrieval to facilitate precise 3D layout specification. Diverging from existing methods that lack control and precision, our approach allows for detailed customization of room-scale environments. To overcome the limitations of previous methods, we introduce an adaptive viewpoint selection algorithm that allows the system to generate the furniture texture and geometry from more favorable views than predefined camera trajectories. Additionally, we employ non-rigid depth registration to ensure alignment between generated objects and their corresponding primitive while allowing for shape variations to maintain diversity. Our method not only enhances the accuracy and aesthetic appeal of generated 3D scenes but also provides a user-friendly platform for detailed room design.

Pan Ji, Yuan Tian, Qingan Yan et al.

We present a robust visual-inertial SLAM system that combines the benefits of Convolutional Neural Networks (CNNs) and planar constraints. Our system leverages a CNN to predict the depth map and the corresponding uncertainty map for each image. The CNN depth effectively bootstraps the back-end optimization of SLAM and meanwhile the CNN uncertainty adaptively weighs the contribution of each feature point to the back-end optimization. Given the gravity direction from the inertial sensor, we further present a fast plane detection method that detects horizontal planes via one-point RANSAC and vertical planes via two-point RANSAC. Those stably detected planes are in turn used to regularize the back-end optimization of SLAM. We evaluate our system on a public dataset, \ie, EuRoC, and demonstrate improved results over a state-of-the-art SLAM system, \ie, ORB-SLAM3.

Jiayu Yang, Taizhang Shang, Weixuan Sun et al.

This report presents a comprehensive framework for generating high-quality 3D shapes and textures from diverse input prompts, including single images, multi-view images, and text descriptions. The framework consists of 3D shape generation and texture generation. (1). The 3D shape generation pipeline employs a Variational Autoencoder (VAE) to encode implicit 3D geometries into a latent space and a diffusion network to generate latents conditioned on input prompts, with modifications to enhance model capacity. An alternative Artist-Created Mesh (AM) generation approach is also explored, yielding promising results for simpler geometries. (2). Texture generation involves a multi-stage process starting with frontal images generation followed by multi-view images generation, RGB-to-PBR texture conversion, and high-resolution multi-view texture refinement. A consistency scheduler is plugged into every stage, to enforce pixel-wise consistency among multi-view textures during inference, ensuring seamless integration. The pipeline demonstrates effective handling of diverse input formats, leveraging advanced neural architectures and novel methodologies to produce high-quality 3D content. This report details the system architecture, experimental results, and potential future directions to improve and expand the framework. The source code and pretrained weights are released at: https://github.com/Tencent/Tencent-XR-3DGen.

Zhenyue Qin, Yiqun Zhang Saeed Anwar, Dongwoo Kim et al.

Most existing graph neural networks (GNNs) learn node embeddings using the framework of message passing and aggregation. Such GNNs are incapable of learning relative positions between graph nodes within a graph. To empower GNNs with the awareness of node positions, some nodes are set as anchors. Then, using the distances from a node to the anchors, GNNs can infer relative positions between nodes. However, P-GNNs arbitrarily select anchors, leading to compromising position-awareness and feature extraction. To eliminate this compromise, we demonstrate that selecting evenly distributed and asymmetric anchors is essential. On the other hand, we show that choosing anchors that can aggregate embeddings of all the nodes within a graph is NP-complete. Therefore, devising efficient optimal algorithms in a deterministic approach is practically not feasible. To ensure position-awareness and bypass NP-completeness, we propose Position-Sensing Graph Neural Networks (PSGNNs), learning how to choose anchors in a back-propagatable fashion. Experiments verify the effectiveness of PSGNNs against state-of-the-art GNNs, substantially improving performance on various synthetic and real-world graph datasets while enjoying stable scalability. Specifically, PSGNNs on average boost AUC more than 14% for pairwise node classification and 18% for link prediction over the existing state-of-the-art position-aware methods. Our source code is publicly available at: https://github.com/ZhenyueQin/PSGNN.

Yang Liu, Saeed Anwar, Zhenyue Qin et al.

The prevalent convolutional neural network (CNN) based image denoising methods extract features of images to restore the clean ground truth, achieving high denoising accuracy. However, these methods may ignore the underlying distribution of clean images, inducing distortions or artifacts in denoising results. This paper proposes a new perspective to treat image denoising as a distribution learning and disentangling task. Since the noisy image distribution can be viewed as a joint distribution of clean images and noise, the denoised images can be obtained via manipulating the latent representations to the clean counterpart. This paper also provides a distribution learning based denoising framework. Following this framework, we present an invertible denoising network, FDN, without any assumptions on either clean or noise distributions, as well as a distribution disentanglement method. FDN learns the distribution of noisy images, which is different from the previous CNN based discriminative mapping. Experimental results demonstrate FDN's capacity to remove synthetic additive white Gaussian noise (AWGN) on both category-specific and remote sensing images. Furthermore, the performance of FDN surpasses that of previously published methods in real image denoising with fewer parameters and faster speed. Our code is available at: https://github.com/Yang-Liu1082/FDN.git.

Zhenyue Qin, Yang Liu, Pan Ji et al.

Skeleton sequences are lightweight and compact, and thus are ideal candidates for action recognition on edge devices. Recent skeleton-based action recognition methods extract features from 3D joint coordinates as spatial-temporal cues, using these representations in a graph neural network for feature fusion to boost recognition performance. The use of first- and second-order features, i.e., joint and bone representations, has led to high accuracy. Nonetheless, many models are still confused by actions that have similar motion trajectories. To address these issues, we propose fusing higher-order features in the form of angular encoding into modern architectures to robustly capture the relationships between joints and body parts. This simple fusion with popular spatial-temporal graph neural networks achieves new state-of-the-art accuracy in two large benchmarks, including NTU60 and NTU120, while employing fewer parameters and reduced run time. Our source code is publicly available at: https://github.com/ZhenyueQin/Angular-Skeleton-Encoding.

Yang Liu, Zhenyue Qin, Saeed Anwar et al.

Invertible networks have various benefits for image denoising since they are lightweight, information-lossless, and memory-saving during back-propagation. However, applying invertible models to remove noise is challenging because the input is noisy, and the reversed output is clean, following two different distributions. We propose an invertible denoising network, InvDN, to address this challenge. InvDN transforms the noisy input into a low-resolution clean image and a latent representation containing noise. To discard noise and restore the clean image, InvDN replaces the noisy latent representation with another one sampled from a prior distribution during reversion. The denoising performance of InvDN is better than all the existing competitive models, achieving a new state-of-the-art result for the SIDD dataset while enjoying less run time. Moreover, the size of InvDN is far smaller, only having 4.2% of the number of parameters compared to the most recently proposed DANet. Further, via manipulating the noisy latent representation, InvDN is also able to generate noise more similar to the original one. Our code is available at: https://github.com/Yang-Liu1082/InvDN.git.

Han Yan, Mingrui Zhang, Yang Li et al.

We present PhyCAGE, the first approach for physically plausible compositional 3D asset generation from a single image. Given an input image, we first generate consistent multi-view images for components of the assets. These images are then fitted with 3D Gaussian Splatting representations. To ensure that the Gaussians representing objects are physically compatible with each other, we introduce a Physical Simulation-Enhanced Score Distillation Sampling (PSE-SDS) technique to further optimize the positions of the Gaussians. It is achieved by setting the gradient of the SDS loss as the initial velocity of the physical simulation, allowing the simulator to act as a physics-guided optimizer that progressively corrects the Gaussians' positions to a physically compatible state. Experimental results demonstrate that the proposed method can generate physically plausible compositional 3D assets given a single image.

Jingnan Gao, Weizhe Liu, Weixuan Sun et al.

State-of-the-art methods for mesh detailization predominantly utilize Generative Adversarial Networks (GANs) to generate detailed meshes from coarse ones. These methods typically learn a specific style code for each category or similar categories without enforcing geometry supervision across different Levels of Detail (LODs). Consequently, such methods often fail to generalize across a broader range of categories and cannot ensure shape consistency throughout the detailization process. In this paper, we introduce MARS, a novel approach for 3D shape detailization. Our method capitalizes on a novel multi-LOD, multi-category mesh representation to learn shape-consistent mesh representations in latent space across different LODs. We further propose a mesh autoregressive model capable of generating such latent representations through next-LOD token prediction. This approach significantly enhances the realism of the generated shapes. Extensive experiments conducted on the challenging 3D Shape Detailization benchmark demonstrate that our proposed MARS model achieves state-of-the-art performance, surpassing existing methods in both qualitative and quantitative assessments. Notably, the model's capability to generate fine-grained details while preserving the overall shape integrity is particularly commendable.

Zhiyuan Wu, Xibin Song, Senbo Wang et al.

3D object reconstruction from single-view image is a fundamental task in computer vision with wide-ranging applications. Recent advancements in Large Reconstruction Models (LRMs) have shown great promise in leveraging multi-view images generated by 2D diffusion models to extract 3D content. However, challenges remain as 2D diffusion models often struggle to produce dense images with strong multi-view consistency, and LRMs tend to amplify these inconsistencies during the 3D reconstruction process. Addressing these issues is critical for achieving high-quality and efficient 3D reconstruction. In this paper, we present CDI3D, a feed-forward framework designed for efficient, high-quality image-to-3D generation with view interpolation. To tackle the aforementioned challenges, we propose to integrate 2D diffusion-based view interpolation into the LRM pipeline to enhance the quality and consistency of the generated mesh. Specifically, our approach introduces a Dense View Interpolation (DVI) module, which synthesizes interpolated images between main views generated by the 2D diffusion model, effectively densifying the input views with better multi-view consistency. We also design a tilt camera pose trajectory to capture views with different elevations and perspectives. Subsequently, we employ a tri-plane-based mesh reconstruction strategy to extract robust tokens from these interpolated and original views, enabling the generation of high-quality 3D meshes with superior texture and geometry. Extensive experiments demonstrate that our method significantly outperforms previous state-of-the-art approaches across various benchmarks, producing 3D content with enhanced texture fidelity and geometric accuracy.

Zhongjin Luo, Yang Li, Mingrui Zhang et al.

While recent advancements have shown remarkable progress in general 3D shape generation models, the challenge of leveraging these approaches to automatically generate wearable 3D assets remains unexplored. To this end, we present BAG, a Body-aligned Asset Generation method to output 3D wearable asset that can be automatically dressed on given 3D human bodies. This is achived by controlling the 3D generation process using human body shape and pose information. Specifically, we first build a general single-image to consistent multiview image diffusion model, and train it on the large Objaverse dataset to achieve diversity and generalizability. Then we train a Controlnet to guide the multiview generator to produce body-aligned multiview images. The control signal utilizes the multiview 2D projections of the target human body, where pixel values represent the XYZ coordinates of the body surface in a canonical space. The body-conditioned multiview diffusion generates body-aligned multiview images, which are then fed into a native 3D diffusion model to produce the 3D shape of the asset. Finally, by recovering the similarity transformation using multiview silhouette supervision and addressing asset-body penetration with physics simulators, the 3D asset can be accurately fitted onto the target human body. Experimental results demonstrate significant advantages over existing methods in terms of image prompt-following capability, shape diversity, and shape quality. Our project page is available at https://bag-3d.github.io/.

Pan Ji, Runze Li, Bir Bhanu et al.

Self-supervised depth estimation for indoor environments is more challenging than its outdoor counterpart in at least the following two aspects: (i) the depth range of indoor sequences varies a lot across different frames, making it difficult for the depth network to induce consistent depth cues, whereas the maximum distance in outdoor scenes mostly stays the same as the camera usually sees the sky; (ii) the indoor sequences contain much more rotational motions, which cause difficulties for the pose network, while the motions of outdoor sequences are pre-dominantly translational, especially for driving datasets such as KITTI. In this paper, special considerations are given to those challenges and a set of good practices are consolidated for improving the performance of self-supervised monocular depth estimation in indoor environments. The proposed method mainly consists of two novel modules, \ie, a depth factorization module and a residual pose estimation module, each of which is designed to respectively tackle the aforementioned challenges. The effectiveness of each module is shown through a carefully conducted ablation study and the demonstration of the state-of-the-art performance on three indoor datasets, \ie, EuRoC, NYUv2, and 7-scenes.

Ze Ma, Yifan Yao, Pan Ji et al.

Estimating 3D human pose and shape from a single image is highly under-constrained. To address this ambiguity, we propose a novel prior, namely kinematic dictionary, which explicitly regularizes the solution space of relative 3D rotations of human joints in the kinematic tree. Integrated with a statistical human model and a deep neural network, our method achieves end-to-end 3D reconstruction without the need of using any shape annotations during the training of neural networks. The kinematic dictionary bridges the gap between in-the-wild images and 3D datasets, and thus facilitates end-to-end training across all types of datasets. The proposed method achieves competitive results on large-scale datasets including Human3.6M, MPI-INF-3DHP, and LSP, while running in real-time given the human bounding boxes.

Pengfei Fang, Pan Ji, Lars Petersson et al.

Modern video person re-identification (re-ID) machines are often trained using a metric learning approach, supervised by a triplet loss. The triplet loss used in video re-ID is usually based on so-called clip features, each aggregated from a few frame features. In this paper, we propose to model the video clip as a set and instead study the distance between sets in the corresponding triplet loss. In contrast to the distance between clip representations, the distance between clip sets considers the pair-wise similarity of each element (i.e., frame representation) between two sets. This allows the network to directly optimize the feature representation at a frame level. Apart from the commonly-used set distance metrics (e.g., ordinary distance and Hausdorff distance), we further propose a hybrid distance metric, tailored for the set-aware triplet loss. Also, we propose a hard positive set construction strategy using the learned class prototypes in a batch. Our proposed method achieves state-of-the-art results across several standard benchmarks, demonstrating the advantages of the proposed method.

Jianyuan Wang, Yiran Zhong, Yuchao Dai et al.

Learning matching costs has been shown to be critical to the success of the state-of-the-art deep stereo matching methods, in which 3D convolutions are applied on a 4D feature volume to learn a 3D cost volume. However, this mechanism has never been employed for the optical flow task. This is mainly due to the significantly increased search dimension in the case of optical flow computation, ie, a straightforward extension would require dense 4D convolutions in order to process a 5D feature volume, which is computationally prohibitive. This paper proposes a novel solution that is able to bypass the requirement of building a 5D feature volume while still allowing the network to learn suitable matching costs from data. Our key innovation is to decouple the connection between 2D displacements and learn the matching costs at each 2D displacement hypothesis independently, ie, displacement-invariant cost learning. Specifically, we apply the same 2D convolution-based matching net independently on each 2D displacement hypothesis to learn a 4D cost volume. Moreover, we propose a displacement-aware projection layer to scale the learned cost volume, which reconsiders the correlation between different displacement candidates and mitigates the multi-modal problem in the learned cost volume. The cost volume is then projected to optical flow estimation through a 2D soft-argmin layer. Extensive experiments show that our approach achieves state-of-the-art accuracy on various datasets, and outperforms all published optical flow methods on the Sintel benchmark.

Pengfei Fang, Pan Ji, Jieming Zhou et al.

Full attention, which generates an attention value per element of the input feature maps, has been successfully demonstrated to be beneficial in visual tasks. In this work, we propose a fully attentional network, termed {\it channel recurrent attention network}, for the task of video pedestrian retrieval. The main attention unit, \textit{channel recurrent attention}, identifies attention maps at the frame level by jointly leveraging spatial and channel patterns via a recurrent neural network. This channel recurrent attention is designed to build a global receptive field by recurrently receiving and learning the spatial vectors. Then, a \textit{set aggregation} cell is employed to generate a compact video representation. Empirical experimental results demonstrate the superior performance of the proposed deep network, outperforming current state-of-the-art results across standard video person retrieval benchmarks, and a thorough ablation study shows the effectiveness of the proposed units.

Ming Zhu, Chao Ma, Pan Ji et al.

In this paper, we focus on exploring the fusion of images and point clouds for 3D object detection in view of the complementary nature of the two modalities, i.e., images possess more semantic information while point clouds specialize in distance sensing. To this end, we present a novel two-stage multi-modal fusion network for 3D object detection, taking both binocular images and raw point clouds as input. The whole architecture facilitates two-stage fusion. The first stage aims at producing 3D proposals through sparse point-wise feature fusion. Within the first stage, we further exploit a joint anchor mechanism that enables the network to utilize 2D-3D classification and regression simultaneously for better proposal generation. The second stage works on the 2D and 3D proposal regions and fuses their dense features. In addition, we propose to use pseudo LiDAR points from stereo matching as a data augmentation method to densify the LiDAR points, as we observe that objects missed by the detection network mostly have too few points especially for far-away objects. Our experiments on the KITTI dataset show that the proposed multi-stage fusion helps the network to learn better representations.

Yuliang Zou, Pan Ji, Quoc-Huy Tran et al.

Monocular visual odometry (VO) suffers severely from error accumulation during frame-to-frame pose estimation. In this paper, we present a self-supervised learning method for VO with special consideration for consistency over longer sequences. To this end, we model the long-term dependency in pose prediction using a pose network that features a two-layer convolutional LSTM module. We train the networks with purely self-supervised losses, including a cycle consistency loss that mimics the loop closure module in geometric VO. Inspired by prior geometric systems, we allow the networks to see beyond a small temporal window during training, through a novel a loss that incorporates temporally distant (e.g., O(100)) frames. Given GPU memory constraints, we propose a stage-wise training mechanism, where the first stage operates in a local time window and the second stage refines the poses with a "global" loss given the first stage features. We demonstrate competitive results on several standard VO datasets, including KITTI and TUM RGB-D.

Buyu Liu, Bingbing Zhuang, Samuel Schulter et al.

In this paper, we address the problem of inferring the layout of complex road scenes from video sequences. To this end, we formulate it as a top-view road attributes prediction problem and our goal is to predict these attributes for each frame both accurately and consistently. In contrast to prior work, we exploit the following three novel aspects: leveraging camera motions in videos, including context cuesand incorporating long-term video information. Specifically, we introduce a model that aims to enforce prediction consistency in videos. Our model consists of one LSTM and one Feature Transform Module (FTM). The former implicitly incorporates the consistency constraint with its hidden states, and the latter explicitly takes the camera motion into consideration when aggregating information along videos. Moreover, we propose to incorporate context information by introducing road participants, e.g. objects, into our model. When the entire video sequence is available, our model is also able to encode both local and global cues, e.g. information from both past and future frames. Experiments on two data sets show that: (1) Incorporating either globalor contextual cues improves the prediction accuracy and leveraging both gives the best performance. (2) Introducing the LSTM and FTM modules improves the prediction consistency in videos. (3) The proposed method outperforms the SOTA by a large margin.

Lokender Tiwari, Pan Ji, Quoc-Huy Tran et al.

Classical monocular Simultaneous Localization And Mapping (SLAM) and the recently emerging convolutional neural networks (CNNs) for monocular depth prediction represent two largely disjoint approaches towards building a 3D map of the surrounding environment. In this paper, we demonstrate that the coupling of these two by leveraging the strengths of each mitigates the other's shortcomings. Specifically, we propose a joint narrow and wide baseline based self-improving framework, where on the one hand the CNN-predicted depth is leveraged to perform pseudo RGB-D feature-based SLAM, leading to better accuracy and robustness than the monocular RGB SLAM baseline. On the other hand, the bundle-adjusted 3D scene structures and camera poses from the more principled geometric SLAM are injected back into the depth network through novel wide baseline losses proposed for improving the depth prediction network, which then continues to contribute towards better pose and 3D structure estimation in the next iteration. We emphasize that our framework only requires unlabeled monocular videos in both training and inference stages, and yet is able to outperform state-of-the-art self-supervised monocular and stereo depth prediction networks (e.g, Monodepth2) and feature-based monocular SLAM system (i.e, ORB-SLAM). Extensive experiments on KITTI and TUM RGB-D datasets verify the superiority of our self-improving geometry-CNN framework.

Bingbing Zhuang, Quoc-Huy Tran, Pan Ji et al.

Self-calibration of camera intrinsics and radial distortion has a long history of research in the computer vision community. However, it remains rare to see real applications of such techniques to modern Simultaneous Localization And Mapping (SLAM) systems, especially in driving scenarios. In this paper, we revisit the geometric approach to this problem, and provide a theoretical proof that explicitly shows the ambiguity between radial distortion and scene depth when two-view geometry is used to self-calibrate the radial distortion. In view of such geometric degeneracy, we propose a learning approach that trains a convolutional neural network (CNN) on a large amount of synthetic data. We demonstrate the utility of our proposed method by applying it as a checkerboard-free calibration tool for SLAM, achieving comparable or superior performance to previous learning and hand-crafted methods.

Tong Zhang, Pan Ji, Mehrtash Harandi et al.

We introduce the Neural Collaborative Subspace Clustering, a neural model that discovers clusters of data points drawn from a union of low-dimensional subspaces. In contrast to previous attempts, our model runs without the aid of spectral clustering. This makes our algorithm one of the kinds that can gracefully scale to large datasets. At its heart, our neural model benefits from a classifier which determines whether a pair of points lies on the same subspace or not. Essential to our model is the construction of two affinity matrices, one from the classifier and the other from a notion of subspace self-expressiveness, to supervise training in a collaborative scheme. We thoroughly assess and contrast the performance of our model against various state-of-the-art clustering algorithms including deep subspace-based ones.

Xuelian Cheng, Yiran Zhong, Yuchao Dai et al.

In this paper, we present LidarStereoNet, the first unsupervised Lidar-stereo fusion network, which can be trained in an end-to-end manner without the need of ground truth depth maps. By introducing a novel "Feedback Loop'' to connect the network input with output, LidarStereoNet could tackle both noisy Lidar points and misalignment between sensors that have been ignored in existing Lidar-stereo fusion studies. Besides, we propose to incorporate a piecewise planar model into network learning to further constrain depths to conform to the underlying 3D geometry. Extensive quantitative and qualitative evaluations on both real and synthetic datasets demonstrate the superiority of our method, which outperforms state-of-the-art stereo matching, depth completion and Lidar-Stereo fusion approaches significantly.

Yiran Zhong, Pan Ji, Jianyuan Wang et al.

Unsupervised deep learning for optical flow computation has achieved promising results. Most existing deep-net based methods rely on image brightness consistency and local smoothness constraint to train the networks. Their performance degrades at regions where repetitive textures or occlusions occur. In this paper, we propose Deep Epipolar Flow, an unsupervised optical flow method which incorporates global geometric constraints into network learning. In particular, we investigate multiple ways of enforcing the epipolar constraint in flow estimation. To alleviate a "chicken-and-egg" type of problem encountered in dynamic scenes where multiple motions may be present, we propose a low-rank constraint as well as a union-of-subspaces constraint for training. Experimental results on various benchmarking datasets show that our method achieves competitive performance compared with supervised methods and outperforms state-of-the-art unsupervised deep-learning methods.

Tong Zhang, Pan Ji, Mehrtash Harandi et al.

Subspace clustering algorithms are notorious for their scalability issues because building and processing large affinity matrices are demanding. In this paper, we introduce a method that simultaneously learns an embedding space along subspaces within it to minimize a notion of reconstruction error, thus addressing the problem of subspace clustering in an end-to-end learning paradigm. To achieve our goal, we propose a scheme to update subspaces within a deep neural network. This in turn frees us from the need of having an affinity matrix to perform clustering. Unlike previous attempts, our method can easily scale up to large datasets, making it unique in the context of unsupervised learning with deep architectures. Our experiments show that our method significantly improves the clustering accuracy while enjoying cheaper memory footprints.

Pan Ji, Tong Zhang, Hongdong Li et al.

We present a novel deep neural network architecture for unsupervised subspace clustering. This architecture is built upon deep auto-encoders, which non-linearly map the input data into a latent space. Our key idea is to introduce a novel self-expressive layer between the encoder and the decoder to mimic the "self-expressiveness" property that has proven effective in traditional subspace clustering. Being differentiable, our new self-expressive layer provides a simple but effective way to learn pairwise affinities between all data points through a standard back-propagation procedure. Being nonlinear, our neural-network based method is able to cluster data points having complex (often nonlinear) structures. We further propose pre-training and fine-tuning strategies that let us effectively learn the parameters of our subspace clustering networks. Our experiments show that the proposed method significantly outperforms the state-of-the-art unsupervised subspace clustering methods.

Pan Ji, Hongdong Li, Yuchao Dai et al.

Rigid structure-from-motion (RSfM) and non-rigid structure-from-motion (NRSfM) have long been treated in the literature as separate (different) problems. Inspired by a previous work which solved directly for 3D scene structure by factoring the relative camera poses out, we revisit the principle of "maximizing rigidity" in structure-from-motion literature, and develop a unified theory which is applicable to both rigid and non-rigid structure reconstruction in a rigidity-agnostic way. We formulate these problems as a convex semi-definite program, imposing constraints that seek to apply the principle of minimizing non-rigidity. Our results demonstrate the efficacy of the approach, with state-of-the-art accuracy on various 3D reconstruction problems.

Pan Ji, Ian Reid, Ravi Garg et al.

In this paper, we present a kernel subspace clustering method that can handle non-linear models. In contrast to recent kernel subspace clustering methods which use predefined kernels, we propose to learn a low-rank kernel matrix, with which mapped data in feature space are not only low-rank but also self-expressive. In this manner, the low-dimensional subspace structures of the (implicitly) mapped data are retained and manifested in the high-dimensional feature space. We evaluate the proposed method extensively on both motion segmentation and image clustering benchmarks, and obtain superior results, outperforming the kernel subspace clustering method that uses standard kernels[Patel 2014] and other state-of-the-art linear subspace clustering methods.

Pan Ji, Hongdong Li, Mathieu Salzmann et al.

Feature tracking is a fundamental problem in computer vision, with applications in many computer vision tasks, such as visual SLAM and action recognition. This paper introduces a novel multi-body feature tracker that exploits a multi-body rigidity assumption to improve tracking robustness under a general perspective camera model. A conventional approach to addressing this problem would consist of alternating between solving two subtasks: motion segmentation and feature tracking under rigidity constraints for each segment. This approach, however, requires knowing the number of motions, as well as assigning points to motion groups, which is typically sensitive to the motion estimates. By contrast, here, we introduce a segmentation-free solution to multi-body feature tracking that bypasses the motion assignment step and reduces to solving a series of subproblems with closed-form solutions. Our experiments demonstrate the benefits of our approach in terms of tracking accuracy and robustness to noise.

Pan Ji, Mathieu Salzmann, Hongdong Li

The Shape Interaction Matrix (SIM) is one of the earliest approaches to performing subspace clustering (i.e., separating points drawn from a union of subspaces). In this paper, we revisit the SIM and reveal its connections to several recent subspace clustering methods. Our analysis lets us derive a simple, yet effective algorithm to robustify the SIM and make it applicable to realistic scenarios where the data is corrupted by noise. We justify our method by intuitive examples and the matrix perturbation theory. We then show how this approach can be extended to handle missing data, thus yielding an efficient and general subspace clustering algorithm. We demonstrate the benefits of our approach over state-of-the-art subspace clustering methods on several challenging motion segmentation and face clustering problems, where the data includes corrupted and missing measurements.