Zhenyang Li

Fanyue Wei, Wei Zeng, Zhenyang Li et al.

Personalized text-to-image models allow users to generate varied styles of images (specified with a sentence) for an object (specified with a set of reference images). While remarkable results have been achieved using diffusion-based generation models, the visual structure and details of the object are often unexpectedly changed during the diffusion process. One major reason is that these diffusion-based approaches typically adopt a simple reconstruction objective during training, which can hardly enforce appropriate structural consistency between the generated and the reference images. To this end, in this paper, we design a novel reinforcement learning framework by utilizing the deterministic policy gradient method for personalized text-to-image generation, with which various objectives, differential or even non-differential, can be easily incorporated to supervise the diffusion models to improve the quality of the generated images. Experimental results on personalized text-to-image generation benchmark datasets demonstrate that our proposed approach outperforms existing state-of-the-art methods by a large margin on visual fidelity while maintaining text-alignment. Our code is available at: \url{https://github.com/wfanyue/DPG-T2I-Personalization}.

Boming Zhao, Bangbang Yang, Zhenyang Li et al.

Expanding an existing tourist photo from a partially captured scene to a full scene is one of the desired experiences for photography applications. Although photo extrapolation has been well studied, it is much more challenging to extrapolate a photo (i.e., selfie) from a narrow field of view to a wider one while maintaining a similar visual style. In this paper, we propose a factorized neural re-rendering model to produce photorealistic novel views from cluttered outdoor Internet photo collections, which enables the applications including controllable scene re-rendering, photo extrapolation and even extrapolated 3D photo generation. Specifically, we first develop a novel factorized re-rendering pipeline to handle the ambiguity in the decomposition of geometry, appearance and illumination. We also propose a composited training strategy to tackle the unexpected occlusion in Internet images. Moreover, to enhance photo-realism when extrapolating tourist photographs, we propose a novel realism augmentation process to complement appearance details, which automatically propagates the texture details from a narrow captured photo to the extrapolated neural rendered image. The experiments and photo editing examples on outdoor scenes demonstrate the superior performance of our proposed method in both photo-realism and downstream applications.

Zhenyang Li, Yangyang Guo, Kejie Wang et al.

Visual Commonsense Reasoning (VCR) remains a significant yet challenging research problem in the realm of visual reasoning. A VCR model generally aims at answering a textual question regarding an image, followed by the rationale prediction for the preceding answering process. Though these two processes are sequential and intertwined, existing methods always consider them as two independent matching-based instances. They, therefore, ignore the pivotal relationship between the two processes, leading to sub-optimal model performance. This paper presents a novel visual attention alignment method to efficaciously handle these two processes in a unified framework. To achieve this, we first design a re-attention module for aggregating the vision attention map produced in each process. Thereafter, the resultant two sets of attention maps are carefully aligned to guide the two processes to make decisions based on the same image regions. We apply this method to both conventional attention and the recent Transformer models and carry out extensive experiments on the VCR benchmark dataset. The results demonstrate that with the attention alignment module, our method achieves a considerable improvement over the baseline methods, evidently revealing the feasibility of the coupling of the two processes as well as the effectiveness of the proposed method.

Mingqing Wang, Jiawei Li, Zhenyang Li et al.

Unsupervised anomaly detection (UAD) has been widely implemented in industrial and medical applications, which reduces the cost of manual annotation and improves efficiency in disease diagnosis. Recently, deep auto-encoder with its variants has demonstrated its advantages in many UAD scenarios. Training on the normal data, these models are expected to locate anomalies by producing higher reconstruction error for the abnormal areas than the normal ones. However, this assumption does not always hold because of the uncontrollable generalization capability. To solve this problem, we present LSGS, a method that builds on Vector Quantised-Variational Autoencoder (VQVAE) with a novel aggregated codebook and transformers with global attention. In this work, the VQVAE focus on feature extraction and reconstruction of images, and the transformers fit the manifold and locate anomalies in the latent space. Then, leveraging the generated encoding sequences that conform to a normal distribution, we can reconstruct a more accurate image for locating the anomalies. Experiments on various datasets demonstrate the effectiveness of the proposed method.

Zhaoyang Xu, Xusheng He, Wei Liu et al.

Temporal-logic video question answering requires a model to reason about when actions occur relative to one another, such as before, after, until, since, overlap, and multi-event chains, rather than merely what is present in a video. Standard vision-language models typically answer such questions in a single pass over a fixed, uniformly sampled set of frames, which is poorly matched to evidence that is often localized to narrow action boundaries or dispersed across several distant events. We present an evidence-seeking agent that treats temporal-logic VideoQA as active exploration. The agent follows a Think-Act-Observe loop driven by a multi-granular sampling toolkit, where every observation is interleaved with its absolute timestamp so that temporal relations reduce to numerical comparisons on a shared time axis. Its behavior is shaped by benchmark structure: a lightweight classifier routes each question to a temporal category, each with a tailored policy, iteration depth, and prompt, while sampling budgets adapt to corpus characteristics and clip length. The resulting training-free system couples Gemini 3.1 Pro with a temporal-reasoning policy and achieves 77.13 AvgAcc on the official TimeLogic test set.

Yizhou Zhao, Zhenyang Li, Xun Guo et al.

Temporal modeling is crucial for various video learning tasks. Most recent approaches employ either factorized (2D+1D) or joint (3D) spatial-temporal operations to extract temporal contexts from the input frames. While the former is more efficient in computation, the latter often obtains better performance. In this paper, we attribute this to a dilemma between the sufficiency and the efficiency of interactions among various positions in different frames. These interactions affect the extraction of task-relevant information shared among frames. To resolve this issue, we prove that frame-by-frame alignments have the potential to increase the mutual information between frame representations, thereby including more task-relevant information to boost effectiveness. Then we propose Alignment-guided Temporal Attention (ATA) to extend 1-dimensional temporal attention with parameter-free patch-level alignments between neighboring frames. It can act as a general plug-in for image backbones to conduct the action recognition task without any model-specific design. Extensive experiments on multiple benchmarks demonstrate the superiority and generality of our module.

Yizhou Zhao, Hengwei Bian, Michael Mu et al.

Cryogenic Electron Tomography (CryoET) is a useful imaging technology in structural biology that is hindered by its need for manual annotations, especially in particle picking. Recent works have endeavored to remedy this issue with few-shot learning or contrastive learning techniques. However, supervised training is still inevitable for them. We instead choose to leverage the power of existing 2D foundation models and present a novel, training-free framework, CryoSAM. In addition to prompt-based single-particle instance segmentation, our approach can automatically search for similar features, facilitating full tomogram semantic segmentation with only one prompt. CryoSAM is composed of two major parts: 1) a prompt-based 3D segmentation system that uses prompts to complete single-particle instance segmentation recursively with Cross-Plane Self-Prompting, and 2) a Hierarchical Feature Matching mechanism that efficiently matches relevant features with extracted tomogram features. They collaborate to enable the segmentation of all particles of one category with just one particle-specific prompt. Our experiments show that CryoSAM outperforms existing works by a significant margin and requires even fewer annotations in particle picking. Further visualizations demonstrate its ability when dealing with full tomogram segmentation for various subcellular structures. Our code is available at: https://github.com/xulabs/aitom

Zefeng Zhang, Xiangzhao Hao, Hengzhu Tang et al.

Visual Spatial Reasoning is crucial for enabling Multimodal Large Language Models (MLLMs) to understand object properties and spatial relationships, yet current models still struggle with 3D-aware reasoning. Existing approaches typically enhance either perception, by augmenting RGB inputs with auxiliary modalities such as depth and segmentation, or reasoning, by training on spatial VQA datasets and applying reinforcement learning, and thus treat these two aspects in isolation. In this work, we investigate whether a unified MLLM can develop an intrinsic ability to enhance spatial perception and, through adaptive interleaved reasoning, achieve stronger spatial intelligence. We propose \textbf{COOPER}, a unified MLLM that leverages depth and segmentation as auxiliary modalities and is trained in two stages to acquire auxiliary modality generation and adaptive, interleaved reasoning capabilities. COOPER achieves an average \textbf{6.91\%} improvement in spatial reasoning while maintaining general performance. Moreover, even a variant trained only for auxiliary modality generation attains a \textbf{7.92\%} gain on distance and size estimation, suggesting that learning to generate auxiliary modalities helps internalize spatial knowledge and strengthen spatial understanding.

Yikang Ding, Zhenyang Li, Dihe Huang et al.

Learning-based multi-view stereo (MVS) methods have made impressive progress and surpassed traditional methods in recent years. However, their accuracy and completeness are still struggling. In this paper, we propose a new method to enhance the performance of existing networks inspired by contrastive learning and feature matching. First, we propose a Contrast Matching Loss (CML), which treats the correct matching points in depth-dimension as positive sample and other points as negative samples, and computes the contrastive loss based on the similarity of features. We further propose a Weighted Focal Loss (WFL) for better classification capability, which weakens the contribution of low-confidence pixels in unimportant areas to the loss according to predicted confidence. Extensive experiments performed on DTU, Tanks and Temples and BlendedMVS datasets show our method achieves state-of-the-art performance and significant improvement over baseline network.

Jun Feng, Zixin Wang, Zhentao Zhang et al.

Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities in visual mathematical reasoning across various existing benchmarks. However, these benchmarks are predominantly based on clean or processed multimodal inputs, without incorporating the images provided by real-world Kindergarten through 12th grade (K-12) educational users. To address this gap, we introduce MathReal, a meticulously curated dataset comprising 2,000 mathematical questions with images captured by handheld mobile devices in authentic scenarios. Each question is an image, containing the question text and visual element. We systematically classify the real images into three primary categories: image quality degradation, perspective variation, and irrelevant content interference, which are further delineated into 14 subcategories. Additionally, MathReal spans five core knowledge and ability categories, which encompass three question types and are divided into three difficulty levels. To comprehensively evaluate the multimodal mathematical reasoning abilities of state-of-the-art MLLMs in real-world scenarios, we design six experimental settings that enable a systematic analysis of their performance. Through extensive experimentation, we find that the problem-solving abilities of existing MLLMs are significantly challenged in realistic educational contexts. Based on this, we conduct a thorough analysis of their performance and error patterns, providing insights into their recognition, comprehension, and reasoning capabilities, and outlining directions for future improvements. Data and code: https://github.com/junfeng0288/MathReal.

Haosen Wang, Zhenyang Li, Yinqiang Zhang et al.

Zero-shot object navigation has advanced rapidly with open-vocabulary detectors, image--text models, and language-guided exploration. However, even after current methods detect a plausible target hypothesis, the agent may still oscillate between exploration and pursuit, or abandon the object near success. We identify this failure mode as an action consistency gap: semantic evidence is repeatedly reinterpreted at each step without persistent commitment across the episode. We introduce ConsistNav, a training-free zero-shot ObjectNav framework built around a semantic executive composed of three coordinated modules: Finite-State Executive Controller stages target pursuit through guarded semantic phases; Persistent Candidate Memory accumulates cross-frame target evidence into stable object hypotheses; and Stability-Aware Action Control suppresses rotational stagnation, ineffective pursuit, and unverified stopping. This design changes neither the detector nor the low-level planner; instead, it controls when semantic evidence should influence navigation and when it should be suppressed or revisited. We conduct extensive experiments on HM3D and MP3D, where ConsistNav achieves state-of-the-art results among compared zero-shot ObjectNav methods and improves SR by 11.4% and SPL by 7.9% over the controlled baseline on MP3D. Ablation studies and real-world deployment experiments further demonstrate the effectiveness and robustness of the proposed executive mechanism.

Weiyou Liu, Zhenyang Li, Weitong Chen

Deep Neural Networks have demonstrated remarkable success in various domains but remain susceptible to adversarial examples, which are slightly altered inputs designed to induce misclassification. While adversarial attacks typically optimize under Lp norm constraints, attacks based on the L0 norm, prioritising input sparsity, are less studied due to their complex and non convex nature. These sparse adversarial examples challenge existing defenses by altering a minimal subset of features, potentially uncovering more subtle DNN weaknesses. However, the current L0 norm attack methodologies face a trade off between accuracy and efficiency either precise but computationally intense or expedient but imprecise. This paper proposes a novel, scalable, and effective approach to generate adversarial examples based on the L0 norm, aimed at refining the robustness evaluation of DNNs against such perturbations.

Huhai Zou, Tianhao Sun, Chuanjiang He et al.

Memory is critical for dialogue agents to maintain coherence and enable continuous adaptation in long-term interactions. While existing memory mechanisms offer basic storage and retrieval capabilities, they are hindered by two primary limitations: (1) rigid memory granularity often disrupts semantic integrity, resulting in fragmented and incoherent memory units; (2) prevalent flat retrieval paradigms rely solely on surface-level semantic similarity, neglecting the structural cues of discourse required to navigate and locate specific episodic contexts. To mitigate these limitations, drawing inspiration from Event Segmentation Theory, we propose ES-Mem, a framework incorporating two core components: (1) a dynamic event segmentation module that partitions long-term interactions into semantically coherent events with distinct boundaries; (2) a hierarchical memory architecture that constructs multi-layered memories and leverages boundary semantics to anchor specific episodic memory for precise context localization. Evaluations on two memory benchmarks demonstrate that ES-Mem yields consistent performance gains over baseline methods. Furthermore, the proposed event segmentation module exhibits robust applicability on dialogue segmentation datasets.

Shaokai He, Kaiwen Wei, Xinyi Zeng et al.

The "reversal curse" refers to the phenomenon where large language models (LLMs) exhibit predominantly unidirectional behavior when processing logically bidirectional relationships. Prior work attributed this to autoregressive training -- predicting the next token inherently favors left-to-right information flow over genuine bidirectional knowledge associations. However, we observe that Diffusion LLMs (DLLMs), despite being trained bidirectionally, also suffer from the reversal curse. To investigate the root causes, we conduct systematic experiments on DLLMs and identify three key reasons: 1) entity fragmentation during training, 2) data asymmetry, and 3) missing entity relations. Motivated by the analysis of these reasons, we propose Diffusion Entity-Relation Modeling (DiffER), which addresses the reversal curse through entity-aware training and balanced data construction. Specifically, DiffER introduces whole-entity masking, which mitigates entity fragmentation by predicting complete entities in a single step. DiffER further employs distribution-symmetric and relation-enhanced data construction strategies to alleviate data asymmetry and missing relations. Extensive experiments demonstrate that DiffER effectively alleviates the reversal curse in Diffusion LLMs, offering new perspectives for future research.

Zefeng Zhang, Hengzhu Tang, Jiawei Sheng et al.

Multimodal Large Language Models excel in various tasks, yet often struggle with modality bias, where the model tends to rely heavily on a single modality and overlook critical information in other modalities, which leads to incorrect focus and generating irrelevant responses. In this paper, we propose using the paradigm of preference optimization to solve the modality bias problem, including RLAIFVBias, a debiased preference optimization dataset, and a Noise Aware Preference Optimization algorithm. Specifically, we first construct the dataset by introducing perturbations to reduce the informational content of certain modalities, compelling the model to rely on a specific modality when generating negative responses. To address the inevitable noise in automatically constructed data, we combine the noise robust Mean Absolute Error with the Binary Cross Entropy in Direct Preference Optimization by a negative Box Cox transformation, and dynamically adjust the algorithm noise robustness based on the evaluated noise levels in the data. Extensive experiments validate our approach, demonstrating not only its effectiveness in mitigating modality bias but also its significant role in minimizing hallucinations.

Zhenyang Li, Zilong Chen, Feifan Qu et al.

In NeRF-aided editing tasks, object movement presents difficulties in supervision generation due to the introduction of variability in object positions. Moreover, the removal operations of certain scene objects often lead to empty regions, presenting challenges for NeRF models in inpainting them effectively. We propose an implicit ray transformation strategy, allowing for direct manipulation of the 3D object's pose by operating on the neural-point in NeRF rays. To address the challenge of inpainting potential empty regions, we present a plug-and-play inpainting module, dubbed differentiable neural-point resampling (DNR), which interpolates those regions in 3D space at the original ray locations within the implicit space, thereby facilitating object removal & scene inpainting tasks. Importantly, employing DNR effectively narrows the gap between ground truth and predicted implicit features, potentially increasing the mutual information (MI) of the features across rays. Then, we leverage DNR and ray transformation to construct a point-based editable NeRF pipeline PR^2T-NeRF. Results primarily evaluated on 3D object removal & inpainting tasks indicate that our pipeline achieves state-of-the-art performance. In addition, our pipeline supports high-quality rendering visualization for diverse editing operations without necessitating extra supervision.

Lutao Jiang, Zidong Cao, Weikai Chen et al.

Promptable instance segmentation is widely adopted in embodied and AR systems, yet the performance of foundation models trained on perspective imagery often degrades on 360° panoramas. In this paper, we introduce Segment Any 4K Panorama (SAP), a foundation model for 4K high-resolution panoramic instance-level segmentation. We reformulate panoramic segmentation as fixed-trajectory perspective video segmentation, decomposing a panorama into overlapping perspective patches sampled along a continuous spherical traversal. This memory-aligned reformulation preserves native 4K resolution while restoring the smooth viewpoint transitions required for stable cross-view propagation. To enable large-scale supervision, we synthesize 183,440 4K-resolution panoramic images with instance segmentation labels using the InfiniGen engine. Trained under this trajectory-aligned paradigm, SAP generalizes effectively to real-world 360° images, achieving +17.2 zero-shot mIoU gain over vanilla SAM2 of different sizes on real-world 4K panorama benchmark.

Kaiwen Wei, Jinpeng Gao, Jiang Zhong et al.

Large language models (LLMs) have shown strong potential in recommendation tasks due to their strengths in language understanding, reasoning and knowledge integration. These capabilities are especially beneficial for review-based recommendation, which relies on semantically rich user-generated texts to reveal fine-grained user preferences and item attributes. However, effectively incorporating reviews into LLM-based recommendation remains challenging due to (1) inefficient to dynamically utilize user reviews under LLMs' constrained context windows, and (2) lacking effective mechanisms to prioritize reviews most relevant to the user's current decision context. To address these challenges, we propose RevBrowse, a review-driven recommendation framework inspired by the "browse-then-decide" decision process commonly observed in online user behavior. RevBrowse integrates user reviews into the LLM-based reranking process to enhance its ability to distinguish between candidate items. To improve the relevance and efficiency of review usage, we introduce PrefRAG, a retrieval-augmented module that disentangles user and item representations into structured forms and adaptively retrieves preference-relevant content conditioned on the target item. Extensive experiments on four Amazon review datasets demonstrate that RevBrowse achieves consistent and significant improvements over strong baselines, highlighting its generalizability and effectiveness in modeling dynamic user preferences. Furthermore, since the retrieval-augmented process is transparent, RevBrowse offers a certain level of interpretability by making visible which reviews influence the final recommendation.

Zhenyang Li, Xiaoyang Bai, Jinfan Lu et al.

Simulating event streams from 3D scenes has become a common practice in event-based vision research, as it meets the demand for large-scale, high temporal frequency data without setting up expensive hardware devices or undertaking extensive data collections. Yet existing methods in this direction typically work with noiseless RGB frames that are costly to render, and therefore they can only achieve a temporal resolution equivalent to 100-300 FPS, far lower than that of real-world event data. In this work, we propose EventTracer, a path tracing-based rendering pipeline that simulates high-fidelity event sequences from complex 3D scenes in an efficient and physics-aware manner. Specifically, we speed up the rendering process via low sample-per-pixel (SPP) path tracing, and train a lightweight event spiking network to denoise the resulting RGB videos into realistic event sequences. To capture the physical properties of event streams, the network is equipped with a bipolar leaky integrate-and-fired (BiLIF) spiking unit and trained with a bidirectional earth mover distance (EMD) loss. Our EventTracer pipeline runs at a speed of about 4 minutes per second of 720p video, and it inherits the merit of accurate spatiotemporal modeling from its path tracing backbone. We show in two downstream tasks that EventTracer captures better scene details and demonstrates a greater similarity to real-world event data than other event simulators, which establishes it as a promising tool for creating large-scale event-RGB datasets at a low cost, narrowing the sim-to-real gap in event-based vision, and boosting various application scenarios such as robotics, autonomous driving, and VRAR.

Yizhou Zhao, Haoyu Chen, Chunjiang Liu et al.

System identification from videos aims to recover object geometry and governing physical laws. Existing methods integrate differentiable rendering with simulation but rely on predefined material priors, limiting their ability to handle unknown ones. We introduce MASIV, the first vision-based framework for material-agnostic system identification. Unlike existing approaches that depend on hand-crafted constitutive laws, MASIV employs learnable neural constitutive models, inferring object dynamics without assuming a scene-specific material prior. However, the absence of full particle state information imposes unique challenges, leading to unstable optimization and physically implausible behaviors. To address this, we introduce dense geometric guidance by reconstructing continuum particle trajectories, providing temporally rich motion constraints beyond sparse visual cues. Comprehensive experiments show that MASIV achieves state-of-the-art performance in geometric accuracy, rendering quality, and generalization ability.

Zhenyang Li, Xiaoyang Bai, Tongchen Zhang et al.

High-fidelity 3D video reconstruction is essential for enabling real-time rendering of dynamic scenes with realistic motion in virtual and augmented reality (VR/AR). The deformation field paradigm of 3D Gaussian splatting has achieved near-photorealistic results in video reconstruction due to the great representation capability of deep deformation networks. However, in videos with complex motion and significant scale variations, deformation networks often overfit to irregular Gaussian trajectories, leading to suboptimal visual quality. Moreover, the gradient-based densification strategy designed for static scene reconstruction proves inadequate to address the absence of dynamic content. In light of these challenges, we propose a flow-empowered velocity field modeling scheme tailored for Gaussian video reconstruction, dubbed FlowGaussian-VR. It consists of two core components: a velocity field rendering (VFR) pipeline which enables optical flow-based optimization, and a flow-assisted adaptive densification (FAD) strategy that adjusts the number and size of Gaussians in dynamic regions. We validate our model's effectiveness on multi-view dynamic reconstruction and novel view synthesis with multiple real-world datasets containing challenging motion scenarios, demonstrating not only notable visual improvements (over 2.5 dB gain in PSNR) and less blurry artifacts in dynamic textures, but also regularized and trackable per-Gaussian trajectories.

Yukun Zhao, Lingyong Yan, Zhenyang Li et al. · baidu

Large language models have achieved remarkable success in various tasks. However, it is challenging for them to learn new tasks incrementally due to catastrophic forgetting. Existing approaches rely on experience replay, optimization constraints, or task differentiation, which encounter strict limitations in real-world scenarios. To address these issues, we propose Joint Flashback Adaptation. We first introduce flashbacks -- a limited number of prompts from old tasks -- when adapting to new tasks and constrain the deviations of the model outputs compared to the original one. We then interpolate latent tasks between flashbacks and new tasks to enable jointly learning relevant latent tasks, new tasks, and flashbacks, alleviating data sparsity in flashbacks and facilitating knowledge sharing for smooth adaptation. Our method requires only a limited number of flashbacks without access to the replay data and is task-agnostic. We conduct extensive experiments on state-of-the-art large language models across 1000+ instruction-following tasks, arithmetic reasoning tasks, and general reasoning tasks. The results demonstrate the superior performance of our method in improving generalization on new tasks and reducing forgetting in old tasks.

Zhenyang Li, Yangyang Guo, Kejie Wang et al.

Visual Commonsense Reasoning (VCR), deemed as one challenging extension of the Visual Question Answering (VQA), endeavors to pursue a more high-level visual comprehension. It is composed of two indispensable processes: question answering over a given image and rationale inference for answer explanation. Over the years, a variety of methods tackling VCR have advanced the performance on the benchmark dataset. Despite significant as these methods are, they often treat the two processes in a separate manner and hence decompose the VCR into two irrelevant VQA instances. As a result, the pivotal connection between question answering and rationale inference is interrupted, rendering existing efforts less faithful on visual reasoning. To empirically study this issue, we perform some in-depth explorations in terms of both language shortcuts and generalization capability to verify the pitfalls of this treatment. Based on our findings, in this paper, we present a plug-and-play knowledge distillation enhanced framework to couple the question answering and rationale inference processes. The key contribution is the introduction of a novel branch, which serves as the bridge to conduct processes connecting. Given that our framework is model-agnostic, we apply it to the existing popular baselines and validate its effectiveness on the benchmark dataset. As detailed in the experimental results, when equipped with our framework, these baselines achieve consistent and significant performance improvements, demonstrating the viability of processes coupling, as well as the superiority of the proposed framework.

Weidong Guo, Mingjun Zhao, Lusheng Zhang et al.

Language model pre-training based on large corpora has achieved tremendous success in terms of constructing enriched contextual representations and has led to significant performance gains on a diverse range of Natural Language Understanding (NLU) tasks. Despite the success, most current pre-trained language models, such as BERT, are trained based on single-grained tokenization, usually with fine-grained characters or sub-words, making it hard for them to learn the precise meaning of coarse-grained words and phrases. In this paper, we propose a simple yet effective pre-training method named LICHEE to efficiently incorporate multi-grained information of input text. Our method can be applied to various pre-trained language models and improve their representation capability. Extensive experiments conducted on CLUE and SuperGLUE demonstrate that our method achieves comprehensive improvements on a wide variety of NLU tasks in both Chinese and English with little extra inference cost incurred, and that our best ensemble model achieves the state-of-the-art performance on CLUE benchmark competition.

Long Chen, Lei Tong, Feixiang Zhou et al.

Underwater image enhancement is such an important vision task due to its significance in marine engineering and aquatic robot. It is usually work as a pre-processing step to improve the performance of high level vision tasks such as underwater object detection. Even though many previous works show the underwater image enhancement algorithms can boost the detection accuracy of the detectors, no work specially focus on investigating the relationship between these two tasks. This is mainly because existing underwater datasets lack either bounding box annotations or high quality reference images, based on which detection accuracy or image quality assessment metrics are calculated. To investigate how the underwater image enhancement methods influence the following underwater object detection tasks, in this paper, we provide a large-scale underwater object detection dataset with both bounding box annotations and high quality reference images, namely OUC dataset. The OUC dataset provides a platform for researchers to comprehensive study the influence of underwater image enhancement algorithms on the underwater object detection task.

Kirill Gavrilyuk, Amir Ghodrati, Zhenyang Li et al.

This paper strives for pixel-level segmentation of actors and their actions in video content. Different from existing works, which all learn to segment from a fixed vocabulary of actor and action pairs, we infer the segmentation from a natural language input sentence. This allows to distinguish between fine-grained actors in the same super-category, identify actor and action instances, and segment pairs that are outside of the actor and action vocabulary. We propose a fully-convolutional model for pixel-level actor and action segmentation using an encoder-decoder architecture optimized for video. To show the potential of actor and action video segmentation from a sentence, we extend two popular actor and action datasets with more than 7,500 natural language descriptions. Experiments demonstrate the quality of the sentence-guided segmentations, the generalization ability of our model, and its advantage for traditional actor and action segmentation compared to the state-of-the-art.

Zhenyang Li, Efstratios Gavves, Mihir Jain et al.

We present a new architecture for end-to-end sequence learning of actions in video, we call VideoLSTM. Rather than adapting the video to the peculiarities of established recurrent or convolutional architectures, we adapt the architecture to fit the requirements of the video medium. Starting from the soft-Attention LSTM, VideoLSTM makes three novel contributions. First, video has a spatial layout. To exploit the spatial correlation we hardwire convolutions in the soft-Attention LSTM architecture. Second, motion not only informs us about the action content, but also guides better the attention towards the relevant spatio-temporal locations. We introduce motion-based attention. And finally, we demonstrate how the attention from VideoLSTM can be used for action localization by relying on just the action class label. Experiments and comparisons on challenging datasets for action classification and localization support our claims.

Roeland De Geest, Efstratios Gavves, Amir Ghodrati et al.

In online action detection, the goal is to detect the start of an action in a video stream as soon as it happens. For instance, if a child is chasing a ball, an autonomous car should recognize what is going on and respond immediately. This is a very challenging problem for four reasons. First, only partial actions are observed. Second, there is a large variability in negative data. Third, the start of the action is unknown, so it is unclear over what time window the information should be integrated. Finally, in real world data, large within-class variability exists. This problem has been addressed before, but only to some extent. Our contributions to online action detection are threefold. First, we introduce a realistic dataset composed of 27 episodes from 6 popular TV series. The dataset spans over 16 hours of footage annotated with 30 action classes, totaling 6,231 action instances. Second, we analyze and compare various baseline methods, showing this is a challenging problem for which none of the methods provides a good solution. Third, we analyze the change in performance when there is a variation in viewpoint, occlusion, truncation, etc. We introduce an evaluation protocol for fair comparison. The dataset, the baselines and the models will all be made publicly available to encourage (much needed) further research on online action detection on realistic data.