Sidan Du

Yifang Xu, Yunzhuo Sun, Yang Li et al.

With the increasing demand for video understanding, video moment and highlight detection (MHD) has emerged as a critical research topic. MHD aims to localize all moments and predict clip-wise saliency scores simultaneously. Despite progress made by existing DETR-based methods, we observe that these methods coarsely fuse features from different modalities, which weakens the temporal intra-modal context and results in insufficient cross-modal interaction. To address this issue, we propose MH-DETR (Moment and Highlight Detection Transformer) tailored for MHD. Specifically, we introduce a simple yet efficient pooling operator within the uni-modal encoder to capture global intra-modal context. Moreover, to obtain temporally aligned cross-modal features, we design a plug-and-play cross-modal interaction module between the encoder and decoder, seamlessly integrating visual and textual features. Comprehensive experiments on QVHighlights, Charades-STA, Activity-Net, and TVSum datasets show that MH-DETR outperforms existing state-of-the-art methods, demonstrating its effectiveness and superiority. Our code is available at https://github.com/YoucanBaby/MH-DETR.

Yifang Xu, Yunzhuo Sun, Zien Xie et al.

Video temporal grounding (VTG) aims to locate specific temporal segments from an untrimmed video based on a linguistic query. Most existing VTG models are trained on extensive annotated video-text pairs, a process that not only introduces human biases from the queries but also incurs significant computational costs. To tackle these challenges, we propose VTG-GPT, a GPT-based method for zero-shot VTG without training or fine-tuning. To reduce prejudice in the original query, we employ Baichuan2 to generate debiased queries. To lessen redundant information in videos, we apply MiniGPT-v2 to transform visual content into more precise captions. Finally, we devise the proposal generator and post-processing to produce accurate segments from debiased queries and image captions. Extensive experiments demonstrate that VTG-GPT significantly outperforms SOTA methods in zero-shot settings and surpasses unsupervised approaches. More notably, it achieves competitive performance comparable to supervised methods. The code is available on https://github.com/YoucanBaby/VTG-GPT

Yifang Xu, Benxiang Zhai, Yunzhuo Sun et al.

Recent advancements in diffusion-based technologies have made significant strides, particularly in identity-preserved portrait generation (IPG). However, when using multiple reference images from the same ID, existing methods typically produce lower-fidelity portraits and struggle to customize face attributes precisely. To address these issues, this paper presents HiFi-Portrait, a high-fidelity method for zero-shot portrait generation. Specifically, we first introduce the face refiner and landmark generator to obtain fine-grained multi-face features and 3D-aware face landmarks. The landmarks include the reference ID and the target attributes. Then, we design HiFi-Net to fuse multi-face features and align them with landmarks, which improves ID fidelity and face control. In addition, we devise an automated pipeline to construct an ID-based dataset for training HiFi-Portrait. Extensive experimental results demonstrate that our method surpasses the SOTA approaches in face similarity and controllability. Furthermore, our method is also compatible with previous SDXL-based works.

Ming Li, Xuejiao Hu, Xueqian Jin et al.

Omnidirectional depth estimation has received much attention from researchers in recent years. However, challenges arise due to camera soiling and variations in camera layouts, affecting the robustness and flexibility of the algorithm. In this paper, we use the geometric constraints and redundant information of multiple 360-degree cameras to achieve robust and flexible multi-view omnidirectional depth estimation. We implement two algorithms, in which the two-stage algorithm obtains initial depth maps by pairwise stereo matching of multiple cameras and fuses the multiple depth maps to achieve the final depth estimation; the one-stage algorithm adopts spherical sweeping based on hypothetical depths to construct a uniform spherical matching cost of the multi-camera images and obtain the depth. Additionally, a generalized epipolar equirectangular projection is introduced to simplify the spherical epipolar constraints. To overcome panorama distortion, a spherical feature extractor is implemented. Furthermore, a synthetic 360-degree dataset consisting of 12K road scene panoramas and 3K ground truth depth maps is presented to train and evaluate 360-degree depth estimation algorithms. Our dataset takes soiled camera lenses and glare into consideration, which is more consistent with the real-world environment. Experiments show that our two algorithms achieve state-of-the-art performance, accurately predicting depth maps even when provided with soiled panorama inputs. The flexibility of the algorithms is experimentally validated in terms of camera layouts and numbers.

Ming Li, Xiong Yang, Chaofan Wu et al.

Omnidirectional depth estimation enables efficient 3D perception over a full 360-degree range. However, in real-world applications such as autonomous driving and robotics, achieving real-time performance and robust cross-scene generalization remains a significant challenge for existing algorithms. In this paper, we propose a real-time omnidirectional depth estimation method for edge computing platforms named Rt-OmniMVS, which introduces the Combined Spherical Sweeping method and implements the lightweight network structure to achieve real-time performance on edge computing platforms. To achieve high accuracy, robustness, and generalization in real-world environments, we introduce a teacher-student learning strategy. We leverage the high-precision stereo matching method as the teacher model to predict pseudo labels for unlabeled real-world data, and utilize data and model augmentation techniques for training to enhance performance of the student model Rt-OmniMVS. We also propose HexaMODE, an omnidirectional depth sensing system based on multi-view fisheye cameras and edge computation device. A large-scale hybrid dataset contains both unlabeled real-world data and synthetic data is collected for model training. Experiments on public datasets demonstrate that proposed method achieves results comparable to state-of-the-art approaches while consuming significantly less resource. The proposed system and algorithm also demonstrate high accuracy in various complex real-world scenarios, both indoors and outdoors, achieving an inference speed of 15 frames per second on edge computing platforms.

Sheng Liu, Yuanzhi Liang, Sidan Du

Recent 3D human motion generation models demonstrate remarkable reconstruction accuracy yet struggle to generalize beyond training distributions. This limitation arises partly from the use of precise 3D supervision, which encourages models to fit fixed coordinate patterns instead of learning the essential 3D structure and motion semantic cues required for robust generalization.To overcome this limitation, we propose Free3D, a framework that synthesizes realistic 3D motions without any 3D motion annotations. Free3D introduces a Motion-Lifting Residual Quantized VAE (ML-RQ) that maps 2D motion sequences into 3D-consistent latent spaces, and a suite of 3D-free regularization objectives enforcing view consistency, orientation coherence, and physical plausibility. Trained entirely on 2D motion data, Free3D generates diverse, temporally coherent, and semantically aligned 3D motions, achieving performance comparable to or even surpassing fully 3D-supervised counterparts. These results suggest that relaxing explicit 3D supervision encourages stronger structural reasoning and generalization, offering a scalable and data-efficient paradigm for 3D motion generation.

Yifang Xu, Chenglei Peng, Ming Li et al.

Deep convolutional neural networks (DCNNs) have achieved great success in monocular depth estimation (MDE). However, few existing works take the contributions for MDE of different levels feature maps into account, leading to inaccurate spatial layout, ambiguous boundaries and discontinuous object surface in the prediction. To better tackle these problems, we propose a Pyramid Feature Attention Network (PFANet) to improve the high-level context features and low-level spatial features. In the proposed PFANet, we design a Dual-scale Channel Attention Module (DCAM) to employ channel attention in different scales, which aggregate global context and local information from the high-level feature maps. To exploit the spatial relationship of visual features, we design a Spatial Pyramid Attention Module (SPAM) which can guide the network attention to multi-scale detailed information in the low-level feature maps. Finally, we introduce scale-invariant gradient loss to increase the penalty on errors in depth-wise discontinuous regions. Experimental results show that our method outperforms state-of-the-art methods on the KITTI dataset.

Yunzhuo Sun, Yifang Xu, Zien Xie et al.

Moment retrieval (MR) and highlight detection (HD) aim to identify relevant moments and highlights in video from corresponding natural language query. Large language models (LLMs) have demonstrated proficiency in various computer vision tasks. However, existing methods for MR\&HD have not yet been integrated with LLMs. In this letter, we propose a novel two-stage model that takes the output of LLMs as the input to the second-stage transformer encoder-decoder. First, MiniGPT-4 is employed to generate the detailed description of the video frame and rewrite the query statement, fed into the encoder as new features. Then, semantic similarity is computed between the generated description and the rewritten queries. Finally, continuous high-similarity video frames are converted into span anchors, serving as prior position information for the decoder. Experiments demonstrate that our approach achieves a state-of-the-art result, and by using only span anchors and similarity scores as outputs, positioning accuracy outperforms traditional methods, like Moment-DETR.

Yifang Xu, Yunzhuo Sun, Benxiang Zhai et al.

Given a video and a linguistic query, video moment retrieval and highlight detection (MR&HD) aim to locate all the relevant spans while simultaneously predicting saliency scores. Most existing methods utilize RGB images as input, overlooking the inherent multi-modal visual signals like optical flow and depth. In this paper, we propose a Multi-modal Fusion and Query Refinement Network (MRNet) to learn complementary information from multi-modal cues. Specifically, we design a multi-modal fusion module to dynamically combine RGB, optical flow, and depth map. Furthermore, to simulate human understanding of sentences, we introduce a query refinement module that merges text at different granularities, containing word-, phrase-, and sentence-wise levels. Comprehensive experiments on QVHighlights and Charades datasets indicate that MRNet outperforms current state-of-the-art methods, achieving notable improvements in MR-mAP@Avg (+3.41) and HD-HIT@1 (+3.46) on QVHighlights.

Yifang Xu, Yunzhuo Sun, Benxiang Zhai et al.

The target of video moment retrieval (VMR) is predicting temporal spans within a video that semantically match a given linguistic query. Existing VMR methods based on multimodal large language models (MLLMs) overly rely on expensive high-quality datasets and time-consuming fine-tuning. Although some recent studies introduce a zero-shot setting to avoid fine-tuning, they overlook inherent language bias in the query, leading to erroneous localization. To tackle the aforementioned challenges, this paper proposes Moment-GPT, a tuning-free pipeline for zero-shot VMR utilizing frozen MLLMs. Specifically, we first employ LLaMA-3 to correct and rephrase the query to mitigate language bias. Subsequently, we design a span generator combined with MiniGPT-v2 to produce candidate spans adaptively. Finally, to leverage the video comprehension capabilities of MLLMs, we apply VideoChatGPT and span scorer to select the most appropriate spans. Our proposed method substantially outperforms the state-ofthe-art MLLM-based and zero-shot models on several public datasets, including QVHighlights, ActivityNet-Captions, and Charades-STA.

Lehuai Xu, Weiming Zhang, Yang Li et al.

Reliable omnidirectional depth estimation from multi-fisheye stereo matching is pivotal to many applications, such as embodied robotics. Existing approaches either rely on spherical sweeping with heuristic fusion strategies to build the cost columns or perform reference-centric stereo matching based on rectified views. However, these methods fail to explicitly exploit geometric relationships between multiple views, rendering them less capable of capturing the global dependencies, visibility, or scale changes. In this paper, we shift to a new perspective and propose a novel reference-free framework, dubbed FreeOmniMVS, via multi-view consistency maximization. The highlight of FreeOmniMVS is that it can aggregate pair-wise correlations into a robust, visibility-aware, and global consensus. As such, it is tolerant to occlusions, partial overlaps, and varying baselines. Specifically, to achieve global coherence, we introduce a novel View-pair Correlation Transformer (VCT) that explicitly models pairwise correlation volumes across all camera view pairs, allowing us to drop unreliable pairs caused by occlusion or out-of-focus observations. To realize scalable and visibility-aware consensus, we propose a lightweight attention mechanism that adaptively fuses the correlation vectors, eliminating the need for a designated reference view and allowing all cameras to contribute equally to the stereo matching process. Extensive experiments on diverse benchmark datasets demonstrate the superiority of our method for globally consistent, visibility-aware, and scale-aware omnidirectional depth estimation.

Sheng Liu, Yuanzhi Liang, Jiepeng Wang et al.

We present Uni-Inter, a unified framework for human motion generation that supports a wide range of interaction scenarios: including human-human, human-object, and human-scene-within a single, task-agnostic architecture. In contrast to existing methods that rely on task-specific designs and exhibit limited generalization, Uni-Inter introduces the Unified Interactive Volume (UIV), a volumetric representation that encodes heterogeneous interactive entities into a shared spatial field. This enables consistent relational reasoning and compound interaction modeling. Motion generation is formulated as joint-wise probabilistic prediction over the UIV, allowing the model to capture fine-grained spatial dependencies and produce coherent, context-aware behaviors. Experiments across three representative interaction tasks demonstrate that Uni-Inter achieves competitive performance and generalizes well to novel combinations of entities. These results suggest that unified modeling of compound interactions offers a promising direction for scalable motion synthesis in complex environments.

Chaofan Wu, Jiaheng Li, Jinghao Cao et al.

Accurate 3D perception is essential for autonomous driving. Traditional methods often struggle with geometric ambiguity due to a lack of geometric prior. To address these challenges, we use omnidirectional depth estimation to introduce geometric prior. Based on the depth information, we propose a Sketch-Coloring framework OmniDepth-Occ. Additionally, our approach introduces a cylindrical voxel representation based on polar coordinate to better align with the radial nature of panoramic camera views. To address the lack of fisheye camera dataset in autonomous driving tasks, we also build a virtual scene dataset with six fisheye cameras, and the data volume has reached twice that of SemanticKITTI. Experimental results demonstrate that our Sketch-Coloring network significantly enhances 3D perception performance.

Zhicheng Cai, Chenglei Peng, Sidan Du

Regularization plays a vital role in machine learning optimization. One novel regularization method called flooding makes the training loss fluctuate around the flooding level. It intends to make the model continue to random walk until it comes to a flat loss landscape to enhance generalization. However, the hyper-parameter flooding level of the flooding method fails to be selected properly and uniformly. We propose a novel method called Jitter to improve it. Jitter is essentially a kind of random loss function. Before training, we randomly sample the Jitter Point from a specific probability distribution. The flooding level should be replaced by Jitter point to obtain a new target function and train the model accordingly. As Jitter point acting as a random factor, we actually add some randomness to the loss function, which is consistent with the fact that there exists innumerable random behaviors in the learning process of the machine learning model and is supposed to make the model more robust. In addition, Jitter performs random walk randomly which divides the loss curve into small intervals and then flipping them over, ideally making the loss curve much flatter and enhancing generalization ability. Moreover, Jitter can be a domain-, task-, and model-independent regularization method and train the model effectively after the training error reduces to zero. Our experimental results show that Jitter method can improve model performance more significantly than the previous flooding method and make the test loss curve descend twice.