ZhiJie Zhang

Miaomiao Cai, Yunshan Ma, Fangqi Zhu et al.

Multi-behavior recommendation improves target-behavior prediction by exploiting heterogeneous auxiliary feedback (e.g., view, collect, and cart), yet its robustness is undermined by behavior-dependent noise and inconsistency. We argue that the key bottleneck is a representation-level failure caused by two coupled heterogeneities. First, intra-behavior representation entanglement arises when multi-hop propagation blends incidental signals with true preferences in the embedding space, making coarse spatial denoising unable to suppress noise without sacrificing informative niche signals. Second, inter-behavior reliability heterogeneity complicates cross-behavior fusion because the predictive value of auxiliary behaviors varies across users and contexts. Without reliability calibration, frequent yet unreliable signals may dominate aggregation and cause target-intent drift. To address this bottleneck, we propose Dynamic Spectral Denoising with Global-Context Attention for Multi-Behavior Recommendation (SpectraMB), a target-oriented model that performs representation purification before reliability-aware fusion. SpectraMB introduces Dynamic Feature-Level Spectral Filtering, which re-parameterizes embeddings along the feature dimension into a feature-frequency space and learns view-adaptive spectral modulation under target supervision, enabling component-wise purification without hand-crafted frequency assumptions. It further proposes Global-Context Attention Fusion, which uses a purified global representation as a context anchor to assess view compatibility and perform reliability-aware aggregation, while a residual global backbone preserves collaborative structure. Extensive experiments on three real-world datasets show that SpectraMB achieves the best results in most evaluation settings and exhibits improved robustness under noisy interactions.

Miaomiao Cai, Zhijie Zhang, Junfeng Fang et al.

Multi-behavior recommendation faces a critical challenge in practice: auxiliary behaviors (e.g., clicks, carts) are often noisy, weakly correlated, or semantically misaligned with the target behavior (e.g., purchase), which leads to biased preference learning and suboptimal performance. While existing methods attempt to fuse these heterogeneous signals, they inherently lack a principled mechanism to ensure robustness against such behavioral inconsistency. In this work, we propose Robust Multi-Behavior Recommendation towards Target Behaviors (RMBRec), a robust multi-behavior recommendation framework grounded in an information-theoretic robustness principle. We interpret robustness as a joint process of maximizing predictive information while minimizing its variance across heterogeneous behavioral environments. Under this perspective, the Representation Robustness Module (RRM) enhances local semantic consistency by maximizing the mutual information between users' auxiliary and target representations, whereas the Optimization Robustness Module (ORM) enforces global stability by minimizing the variance of predictive risks across behaviors, which is an efficient approximation to invariant risk minimization. This local-global collaboration bridges representation purification and optimization invariance in a theoretically coherent way. Extensive experiments on three real-world datasets demonstrate that RMBRec not only outperforms state-of-the-art methods in accuracy but also maintains remarkable stability under various noise perturbations. For reproducibility, our code is available at https://github.com/miaomiao-cai2/RMBRec/.

Zongqi Wan, Zhijie Zhang, Tongyang Li et al.

Multi-arm bandit (MAB) and stochastic linear bandit (SLB) are important models in reinforcement learning, and it is well-known that classical algorithms for bandits with time horizon $T$ suffer $Ω(\sqrt{T})$ regret. In this paper, we study MAB and SLB with quantum reward oracles and propose quantum algorithms for both models with $O(\mbox{poly}(\log T))$ regrets, exponentially improving the dependence in terms of $T$. To the best of our knowledge, this is the first provable quantum speedup for regrets of bandit problems and in general exploitation in reinforcement learning. Compared to previous literature on quantum exploration algorithms for MAB and reinforcement learning, our quantum input model is simpler and only assumes quantum oracles for each individual arm.

Zhiyong Cheng, Yike Jin, Zhijie Zhang et al.

Personalized news recommendation is highly time-sensitive, as user interests are often driven by emerging events, trending topics, and shifting real-world contexts. These dynamics make it essential to model not only users' long-term preferences, which reflect stable reading habits and high-order collaborative patterns, but also their short-term, context-dependent interests that change rapidly over time. However, most existing approaches rely on a single static interaction graph, which struggles to capture both long-term preference patterns and short-term interest changes as user behavior evolves. To address this challenge, we propose a unified framework that learns user preferences from both global and local temporal perspectives. A global preference modeling component captures long-term collaborative signals from the overall interaction graph, while a local preference modeling component partitions historical interactions into stage-wise temporal subgraphs to represent short-term dynamics. Within this module, an LSTM branch models the progressive evolution of recent interests, and a self-attention branch captures long-range temporal dependencies. Extensive experiments on two large-scale real-world datasets show that our approach consistently outperforms strong baselines and delivers fresher and more relevant recommendations across diverse user behaviors and temporal settings.

Rohit Mukherjee, Hannah K. Friedrich, Beth Tellman et al.

Urban flooding affects lives and infrastructure worldwide. Mapping inundation in complex urban environments from satellite imagery remains challenging due to limited spatial resolution, infrequent acquisitions, and cloud cover. We present Urban Flood Observations (UFO), a global, hand-labeled dataset of post-flood inundation in diverse urban settings. UFO comprises 215 image chips (1024 by 1024 pixels) from 14 flood events between 2017 and 2021, derived from 3 m PlanetScope imagery. Each chip is annotated with two classes: 'inundated' (all visible surface water, including floodwater and pre-existing water bodies (permanent or seasonal)) and 'non-inundated'. To demonstrate the dataset's utility, we trained a segmentation model using leave-one-event-out cross-validation, achieving a mean Intersection over Union (IoU) of 77.3. We also used UFO to evaluate two widely used surface water products, the Sentinel-1-based NASA IMPACT model and Google's 10 m Dynamic World water class, which yielded IoUs of 44.1 and 48.1, respectively. UFO is publicly available to support the development and validation of urban inundation mapping methods.

Saurabh Kaushik, Lalit Maurya, Elizabeth Tellman et al.

Geo-Foundational Models (GFMs) enable fast and reliable extraction of spatiotemporal information from satellite imagery, improving flood inundation mapping by leveraging location and time embeddings. Despite their potential, it remains unclear whether GFMs outperform traditional models like U-Net. A systematic comparison across sensors and data availability scenarios is still lacking, which is an essential step to guide end-users in model selection. To address this, we evaluate three GFMs, Prithvi 2.0, Clay V1.5, DOFA, and UViT (a Prithvi variant), against TransNorm, U-Net, and Attention U-Net using PlanetScope, Sentinel-1, and Sentinel-2. We observe competitive performance among all GFMs, with only 2-5% variation between the best and worst models across sensors. Clay outperforms others on PlanetScope (0.79 mIoU) and Sentinel-2 (0.70), while Prithvi leads on Sentinel-1 (0.57). In leave-one-region-out cross-validation across five regions, Clay shows slightly better performance across all sensors (mIoU: 0.72(0.04), 0.66(0.07), 0.51(0.08)) compared to Prithvi (0.70(0.05), 0.64(0.09), 0.49(0.13)) and DOFA (0.67(0.07), 0.64(0.04), 0.49(0.09)) for PlanetScope, Sentinel-2, and Sentinel-1, respectively. Across all 19 sites, leave-one-region-out cross-validation reveals a 4% improvement by Clay compared to U-Net. Visual inspection highlights Clay's superior ability to retain fine details. Few-shot experiments show Clay achieves 0.64 mIoU on PlanetScope with just five training images, outperforming Prithvi (0.24) and DOFA (0.35). In terms of computational time, Clay is a better choice due to its smaller model size (26M parameters), making it ~3x faster than Prithvi (650M) and 2x faster than DOFA (410M). Contrary to previous findings, our results suggest GFMs offer small to moderate improvements in flood mapping accuracy at lower computational cost and labeling effort compared to traditional U-Net.

Yan Liu, Zhijie Zhang, Li Niu et al.

Object detection has achieved promising success, but requires large-scale fully-annotated data, which is time-consuming and labor-extensive. Therefore, we consider object detection with mixed supervision, which learns novel object categories using weak annotations with the help of full annotations of existing base object categories. Previous works using mixed supervision mainly learn the class-agnostic objectness from fully-annotated categories, which can be transferred to upgrade the weak annotations to pseudo full annotations for novel categories. In this paper, we further transfer mask prior and semantic similarity to bridge the gap between novel categories and base categories. Specifically, the ability of using mask prior to help detect objects is learned from base categories and transferred to novel categories. Moreover, the semantic similarity between objects learned from base categories is transferred to denoise the pseudo full annotations for novel categories. Experimental results on three benchmark datasets demonstrate the effectiveness of our method over existing methods. Codes are available at https://github.com/bcmi/TraMaS-Weak-Shot-Object-Detection.

Aojun Zhou, Yukun Ma, Junnan Zhu et al.

Sparsity in Deep Neural Networks (DNNs) has been widely studied to compress and accelerate the models on resource-constrained environments. It can be generally categorized into unstructured fine-grained sparsity that zeroes out multiple individual weights distributed across the neural network, and structured coarse-grained sparsity which prunes blocks of sub-networks of a neural network. Fine-grained sparsity can achieve a high compression ratio but is not hardware friendly and hence receives limited speed gains. On the other hand, coarse-grained sparsity cannot concurrently achieve both apparent acceleration on modern GPUs and decent performance. In this paper, we are the first to study training from scratch an N:M fine-grained structured sparse network, which can maintain the advantages of both unstructured fine-grained sparsity and structured coarse-grained sparsity simultaneously on specifically designed GPUs. Specifically, a 2:4 sparse network could achieve 2x speed-up without performance drop on Nvidia A100 GPUs. Furthermore, we propose a novel and effective ingredient, sparse-refined straight-through estimator (SR-STE), to alleviate the negative influence of the approximated gradients computed by vanilla STE during optimization. We also define a metric, Sparse Architecture Divergence (SAD), to measure the sparse network's topology change during the training process. Finally, We justify SR-STE's advantages with SAD and demonstrate the effectiveness of SR-STE by performing comprehensive experiments on various tasks. Source codes and models are available at https://github.com/NM-sparsity/NM-sparsity.

Zhijie Zhang, Huazhu Fu, Hang Dai et al.

Segmentation is a fundamental task in medical image analysis. However, most existing methods focus on primary region extraction and ignore edge information, which is useful for obtaining accurate segmentation. In this paper, we propose a generic medical segmentation method, called Edge-aTtention guidance Network (ET-Net), which embeds edge-attention representations to guide the segmentation network. Specifically, an edge guidance module is utilized to learn the edge-attention representations in the early encoding layers, which are then transferred to the multi-scale decoding layers, fused using a weighted aggregation module. The experimental results on four segmentation tasks (i.e., optic disc/cup and vessel segmentation in retinal images, and lung segmentation in chest X-Ray and CT images) demonstrate that preserving edge-attention representations contributes to the final segmentation accuracy, and our proposed method outperforms current state-of-the-art segmentation methods. The source code of our method is available at https://github.com/ZzzJzzZ/ETNet.

Chang Gong, Wanrui Bian, Zhijie Zhang et al.

Graph computational tasks are inherently challenging and often demand the development of advanced algorithms for effective solutions. With the emergence of large language models (LLMs), researchers have begun investigating their potential to address these tasks. However, existing approaches are constrained by LLMs' limited capability to comprehend complex graph structures and their high inference costs, rendering them impractical for handling large-scale graphs. Inspired by human approaches to graph problems, we introduce a novel framework, PIE (Pseudocode-Injection-Enhanced LLM Reasoning for Graph Computational Tasks), which consists of three key steps: problem understanding, prompt design, and code generation. In this framework, LLMs are tasked with understanding the problem and extracting relevant information to generate correct code. The responsibility for analyzing the graph structure and executing the code is delegated to the interpreter. We inject task-related pseudocodes into the prompts to further assist the LLMs in generating efficient code. We also employ cost-effective trial-and-error techniques to ensure that the LLM-generated code executes correctly. Unlike other methods that require invoking LLMs for each individual test case, PIE only calls the LLM during the code generation phase, allowing the generated code to be reused and significantly reducing inference costs. Extensive experiments demonstrate that PIE outperforms existing baselines in terms of both accuracy and computational efficiency.

Hui Liu, Zhijie Zhang

This paper re-examines the problem of fairly and efficiently allocating indivisible goods among agents with additive bivalued valuations. Garg and Murhekar (2021) proposed a polynomial-time algorithm that purported to find an EFX and fPO allocation. However, we provide a counterexample demonstrating that their algorithm may fail to terminate. To address this issue, we propose a new polynomial-time algorithm that computes a WEFX (Weighted Envy-Free up to any good) and fPO allocation, thereby correcting the prior approach and offering a more general solution. Furthermore, we show that our algorithm can be adapted to compute a WEQX (Weighted Equitable up to any good) and fPO allocation.

Yuxiang Mao, Zhijie Zhang, Zhiheng Zhang et al.

Expressions are fundamental to conveying human emotions. With the rapid advancement of AI-generated content (AIGC), realistic and expressive 3D facial animation has become increasingly crucial. Despite recent progress in speech-driven lip-sync for talking-face animation, generating emotionally expressive talking faces remains underexplored. A major obstacle is the scarcity of real emotional 3D talking-face datasets due to the high cost of data capture. To address this, we model facial animation driven by both speech and emotion as a linear additive problem. Leveraging a 3D talking-face dataset with neutral expressions (VOCAset) and a dataset of 3D expression sequences (Florence4D), we jointly learn a set of blendshapes driven by speech and emotion. We introduce a sparsity constraint loss to encourage disentanglement between the two types of blendshapes while allowing the model to capture inherent secondary cross-domain deformations present in the training data. The learned blendshapes can be further mapped to the expression and jaw pose parameters of the FLAME model, enabling the animation of 3D Gaussian avatars. Qualitative and quantitative experiments demonstrate that our method naturally generates talking faces with specified expressions while maintaining accurate lip synchronization. Perceptual studies further show that our approach achieves superior emotional expressivity compared to existing methods, without compromising lip-sync quality.

Weishi Wang, Hengchang Hu, Zhijie Zhang et al.

Document AI (DAI) has emerged as a vital application area, and is significantly transformed by the advent of large language models (LLMs). While earlier approaches relied on encoder-decoder architectures, decoder-only LLMs have revolutionized DAI, bringing remarkable advancements in understanding and generation. This survey provides a comprehensive overview of DAI's evolution, highlighting current research attempts and future prospects of LLMs in this field. We explore key advancements and challenges in multimodal, multilingual, and retrieval-augmented DAI, while also suggesting future research directions, including agent-based approaches and document-specific foundation models. This paper aims to provide a structured analysis of the state-of-the-art in DAI and its implications for both academic and practical applications.

Yunhao Quan, Chuang Gao, Nan Cheng et al.

In Automatic Modulation Classification (AMC), deep learning methods have shown remarkable performance, offering significant advantages over traditional approaches and demonstrating their vast potential. Nevertheless, notable drawbacks, particularly in their high demands for storage, computational resources, and large-scale labeled data, which limit their practical application in real-world scenarios. To tackle this issue, this paper innovatively proposes an automatic modulation classification model based on the Adaptive Lightweight Wavelet Neural Network (ALWNN) and the few-shot framework (MALWNN). The ALWNN model, by integrating the adaptive wavelet neural network and depth separable convolution, reduces the number of model parameters and computational complexity. The MALWNN framework, using ALWNN as an encoder and incorporating prototype network technology, decreases the model's dependence on the quantity of samples. Simulation results indicate that this model performs remarkably well on mainstream datasets. Moreover, in terms of Floating Point Operations Per Second (FLOPS) and Normalized Multiply - Accumulate Complexity (NMACC), ALWNN significantly reduces computational complexity compared to existing methods. This is further validated by real-world system tests on USRP and Raspberry Pi platforms. Experiments with MALWNN show its superior performance in few-shot learning scenarios compared to other algorithms.

Shi Feng, Nuoya Xiong, Zhijie Zhang et al.

Pseudo log-likelihood is a type of maximum likelihood estimation (MLE) method used in various fields including contextual bandits, influence maximization of social networks, and causal bandits. However, in previous literature \citep{li2017provably, zhang2022online, xiong2022combinatorial, feng2023combinatorial1, feng2023combinatorial2}, the log-likelihood function may not be bounded, which may result in the algorithm they proposed not well-defined. In this paper, we give a counterexample that the maximum pseudo log-likelihood estimation fails and then provide a solution to correct the algorithms in \citep{li2017provably, zhang2022online, xiong2022combinatorial, feng2023combinatorial1, feng2023combinatorial2}.

Zongqi Wan, Jialin Zhang, Wei Chen et al.

We investigate the online bandit learning of the monotone multi-linear DR-submodular functions, designing the algorithm $\mathtt{BanditMLSM}$ that attains $O(T^{2/3}\log T)$ of $(1-1/e)$-regret. Then we reduce submodular bandit with partition matroid constraint and bandit sequential monotone maximization to the online bandit learning of the monotone multi-linear DR-submodular functions, attaining $O(T^{2/3}\log T)$ of $(1-1/e)$-regret in both problems, which improve the existing results. To the best of our knowledge, we are the first to give a sublinear regret algorithm for the submodular bandit with partition matroid constraint. A special case of this problem is studied by Streeter et al.(2009). They prove a $O(T^{4/5})$ $(1-1/e)$-regret upper bound. For the bandit sequential submodular maximization, the existing work proves an $O(T^{2/3})$ regret with a suboptimal $1/2$ approximation ratio (Niazadeh et al. 2021).

Binglu Wang, Tao Hu, Baoshan Li et al.

Gaze object prediction is a newly proposed task that aims to discover the objects being stared at by humans. It is of great application significance but still lacks a unified solution framework. An intuitive solution is to incorporate an object detection branch into an existing gaze prediction method. However, previous gaze prediction methods usually use two different networks to extract features from scene image and head image, which would lead to heavy network architecture and prevent each branch from joint optimization. In this paper, we build a novel framework named GaTector to tackle the gaze object prediction problem in a unified way. Particularly, a specific-general-specific (SGS) feature extractor is firstly proposed to utilize a shared backbone to extract general features for both scene and head images. To better consider the specificity of inputs and tasks, SGS introduces two input-specific blocks before the shared backbone and three task-specific blocks after the shared backbone. Specifically, a novel Defocus layer is designed to generate object-specific features for the object detection task without losing information or requiring extra computations. Moreover, the energy aggregation loss is introduced to guide the gaze heatmap to concentrate on the stared box. In the end, we propose a novel wUoC metric that can reveal the difference between boxes even when they share no overlapping area. Extensive experiments on the GOO dataset verify the superiority of our method in all three tracks, i.e. object detection, gaze estimation, and gaze object prediction.

Zhijie Zhang, Wei Chen, Xiaoming Sun et al.

We study the online influence maximization (OIM) problem in social networks, where the learner repeatedly chooses seed nodes to generate cascades, observes the cascade feedback, and gradually learns the best seeds that generate the largest cascade in multiple rounds. In the demand of the real world, we work with node-level feedback instead of the common edge-level feedback in the literature. The edge-level feedback reveals all edges that pass through information in a cascade, whereas the node-level feedback only reveals the activated nodes with timestamps. The node-level feedback is arguably more realistic since in practice it is relatively easy to observe who is influenced but very difficult to observe from which relationship (edge) the influence comes. Previously, there is a nearly optimal $\tilde{O}(\sqrt{T})$-regret algorithm for OIM problem under the linear threshold (LT) diffusion model with node-level feedback. It remains unknown whether the same algorithm exists for the independent cascade (IC) diffusion model. In this paper, we resolve this open problem by presenting an $\tilde{O}(\sqrt{T})$-regret algorithm for OIM problem under the IC model with node-level feedback.

Yifan Xu, Zhijie Zhang, Mengdan Zhang et al.

Vision transformers (ViTs) have recently received explosive popularity, but the huge computational cost is still a severe issue. Since the computation complexity of ViT is quadratic with respect to the input sequence length, a mainstream paradigm for computation reduction is to reduce the number of tokens. Existing designs include structured spatial compression that uses a progressive shrinking pyramid to reduce the computations of large feature maps, and unstructured token pruning that dynamically drops redundant tokens. However, the limitation of existing token pruning lies in two folds: 1) the incomplete spatial structure caused by pruning is not compatible with structured spatial compression that is commonly used in modern deep-narrow transformers; 2) it usually requires a time-consuming pre-training procedure. To tackle the limitations and expand the applicable scenario of token pruning, we present Evo-ViT, a self-motivated slow-fast token evolution approach for vision transformers. Specifically, we conduct unstructured instance-wise token selection by taking advantage of the simple and effective global class attention that is native to vision transformers. Then, we propose to update the selected informative tokens and uninformative tokens with different computation paths, namely, slow-fast updating. Since slow-fast updating mechanism maintains the spatial structure and information flow, Evo-ViT can accelerate vanilla transformers of both flat and deep-narrow structures from the very beginning of the training process. Experimental results demonstrate that our method significantly reduces the computational cost of vision transformers while maintaining comparable performance on image classification.

Zhijie Zhang, Wei Chen, Xiaoming Sun et al.

Influence maximization is the task of selecting a small number of seed nodes in a social network to maximize the influence spread from these seeds. It has been widely investigated in the past two decades. In the canonical setting, the social network and its diffusion parameters are given as input. In this paper, we consider the more realistic sampling setting where the network is unknown and we only have a set of passively observed cascades that record the sets of activated nodes at each diffusion step. We study the task of influence maximization from these cascade samples (IMS) and present constant approximation algorithms for it under mild conditions on the seed set distribution. To achieve the optimization goal, we also provide a novel solution to the network inference problem, that is, learning diffusion parameters and the network structure from the cascade data. Compared with prior solutions, our network inference algorithms require weaker assumptions and do not rely on maximum-likelihood estimation and convex programming. Our IMS algorithms enhance the learning-and-then-optimization approach by allowing a constant approximation ratio even when the diffusion parameters are hard to learn, and we do not need any assumption related to the network structure or diffusion parameters.

Tian Meng, Yang Tao, Ziqi Chen et al.

Eddy current testing (ECT) is an effective technique in the evaluation of the depth of metal surface defects. However, in practice, the evaluation primarily relies on the experience of an operator and is often carried out by manual inspection. In this paper, we address the challenges of automatic depth evaluation of metal surface defects by virtual of state-of-the-art deep learning (DL) techniques. The main contributions are three-fold. Firstly, a highly-integrated portable ECT device is developed, which takes advantage of an advanced field programmable gate array (Zynq-7020 system on chip) and provides fast data acquisition and in-phase/quadrature demodulation. Secondly, a dataset, termed as MDDECT, is constructed using the ECT device by human operators and made openly available. It contains 48,000 scans from 18 defects of different depths and lift-offs. Thirdly, the depth evaluation problem is formulated as a time series classification problem, and various state-of-the-art 1-d residual convolutional neural networks are trained and evaluated on the MDDECT dataset. A 38-layer 1-d ResNeXt achieves an accuracy of 93.58% in discriminating the surface defects in a stainless steel sheet. The depths of the defects vary from 0.3 mm to 2.0 mm in a resolution of 0.1 mm. In addition, results show that the trained ResNeXt1D-38 model is immune to lift-off signals.

Wei Chen, Xiaoming Sun, Jialin Zhang et al.

We revisit the optimization from samples (OPS) model, which studies the problem of optimizing objective functions directly from the sample data. Previous results showed that we cannot obtain a constant approximation ratio for the maximum coverage problem using polynomially many independent samples of the form $\{S_i, f(S_i)\}_{i=1}^t$ (Balkanski et al., 2017), even if coverage functions are $(1 - ε)$-PMAC learnable using these samples (Badanidiyuru et al., 2012), which means most of the function values can be approximately learned very well with high probability. In this work, to circumvent the impossibility result of OPS, we propose a stronger model called optimization from structured samples (OPSS) for coverage functions, where the data samples encode the structural information of the functions. We show that under three general assumptions on the sample distributions, we can design efficient OPSS algorithms that achieve a constant approximation for the maximum coverage problem. We further prove a constant lower bound under these assumptions, which is tight when not considering computational efficiency. Moreover, we also show that if we remove any one of the three assumptions, OPSS for the maximum coverage problem has no constant approximation.

Wenguan Wang, Zhijie Zhang, Siyuan Qi et al.

This work proposes to combine neural networks with the compositional hierarchy of human bodies for efficient and complete human parsing. We formulate the approach as a neural information fusion framework. Our model assembles the information from three inference processes over the hierarchy: direct inference (directly predicting each part of a human body using image information), bottom-up inference (assembling knowledge from constituent parts), and top-down inference (leveraging context from parent nodes). The bottom-up and top-down inferences explicitly model the compositional and decompositional relations in human bodies, respectively. In addition, the fusion of multi-source information is conditioned on the inputs, i.e., by estimating and considering the confidence of the sources. The whole model is end-to-end differentiable, explicitly modeling information flows and structures. Our approach is extensively evaluated on four popular datasets, outperforming the state-of-the-arts in all cases, with a fast processing speed of 23fps. Our code and results have been released to help ease future research in this direction.