Jie Zhou

Wenliang Zhao, Minglei Shi, Xumin Yu et al.

Building on the success of diffusion models in visual generation, flow-based models reemerge as another prominent family of generative models that have achieved competitive or better performance in terms of both visual quality and inference speed. By learning the velocity field through flow-matching, flow-based models tend to produce a straighter sampling trajectory, which is advantageous during the sampling process. However, unlike diffusion models for which fast samplers are well-developed, efficient sampling of flow-based generative models has been rarely explored. In this paper, we propose a framework called FlowTurbo to accelerate the sampling of flow-based models while still enhancing the sampling quality. Our primary observation is that the velocity predictor's outputs in the flow-based models will become stable during the sampling, enabling the estimation of velocity via a lightweight velocity refiner. Additionally, we introduce several techniques including a pseudo corrector and sample-aware compilation to further reduce inference time. Since FlowTurbo does not change the multi-step sampling paradigm, it can be effectively applied for various tasks such as image editing, inpainting, etc. By integrating FlowTurbo into different flow-based models, we obtain an acceleration ratio of 53.1%$\sim$58.3% on class-conditional generation and 29.8%$\sim$38.5% on text-to-image generation. Notably, FlowTurbo reaches an FID of 2.12 on ImageNet with 100 (ms / img) and FID of 3.93 with 38 (ms / img), achieving the real-time image generation and establishing the new state-of-the-art. Code is available at https://github.com/shiml20/FlowTurbo.

Yiran Yang, Jinchao Zhang, Ying Deng et al. · tencent-ai

Inspired by the success of the text-to-image (T2I) generation task, many researchers are devoting themselves to the text-to-video (T2V) generation task. Most of the T2V frameworks usually inherit from the T2I model and add extra-temporal layers of training to generate dynamic videos, which can be viewed as a fine-tuning task. However, the traditional 3D-Unet is a serial mode and the temporal layers follow the spatial layers, which will result in high GPU memory and training time consumption according to its serial feature flow. We believe that this serial mode will bring more training costs with the large diffusion model and massive datasets, which are not environmentally friendly and not suitable for the development of the T2V. Therefore, we propose a highly efficient spatial-temporal parallel training paradigm for T2V tasks, named Mobius. In our 3D-Unet, the temporal layers and spatial layers are parallel, which optimizes the feature flow and backpropagation. The Mobius will save 24% GPU memory and 12% training time, which can greatly improve the T2V fine-tuning task and provide a novel insight for the AIGC community. We will release our codes in the future.

Yanran Zhang, Ziyi Wang, Wenzhao Zheng et al.

Generating interactive and dynamic 4D scenes from a single static image remains a core challenge. Most existing generate-then-reconstruct and reconstruct-then-generate methods decouple geometry from motion, causing spatiotemporal inconsistencies and poor generalization. To address these, we extend the reconstruct-then-generate framework to jointly perform Motion generation and geometric Reconstruction for 4D Synthesis (MoRe4D). We first introduce TrajScene-60K, a large-scale dataset of 60,000 video samples with dense point trajectories, addressing the scarcity of high-quality 4D scene data. Based on this, we propose a diffusion-based 4D Scene Trajectory Generator (4D-STraG) to jointly generate geometrically consistent and motion-plausible 4D point trajectories. To leverage single-view priors, we design a depth-guided motion normalization strategy and a motion-aware module for effective geometry and dynamics integration. We then propose a 4D View Synthesis Module (4D-ViSM) to render videos with arbitrary camera trajectories from 4D point track representations. Experiments show that MoRe4D generates high-quality 4D scenes with multi-view consistency and rich dynamic details from a single image. Code: https://github.com/Zhangyr2022/MoRe4D.

Dong Zhuo, Wenzhao Zheng, Jiahe Guo et al.

Perceiving and reconstructing 4D spatial-temporal geometry from videos is a fundamental yet challenging computer vision task. To facilitate interactive and real-time applications, we propose a streaming 4D visual geometry transformer that shares a similar philosophy with autoregressive large language models. We explore a simple and efficient design and employ a causal transformer architecture to process the input sequence in an online manner. We use temporal causal attention and cache the historical keys and values as implicit memory to enable efficient streaming long-term 4D reconstruction. This design can handle real-time 4D reconstruction by incrementally integrating historical information while maintaining high-quality spatial consistency. For efficient training, we propose to distill knowledge from the dense bidirectional visual geometry grounded transformer (VGGT) to our causal model. For inference, our model supports the migration of optimized efficient attention operator (e.g., FlashAttention) from the field of large language models. Extensive experiments on various 4D geometry perception benchmarks demonstrate that our model increases the inference speed in online scenarios while maintaining competitive performance, paving the way for scalable and interactive 4D vision systems. Code is available at: https://github.com/wzzheng/StreamVGGT.

Tianyu Yu, Zefan Wang, Chongyi Wang et al. · tsinghua

Multimodal Large Language Models (MLLMs) are undergoing rapid progress and represent the frontier of AI development. However, their training and inference efficiency have emerged as a core bottleneck in making MLLMs more accessible and scalable. To address the challenges, we present MiniCPM-V 4.5, an 8B parameter model designed for high efficiency and strong performance. We introduce three core improvements in model architecture, data strategy and training method: a unified 3D-Resampler model architecture for highly compact encoding over images and videos, a unified learning paradigm for document knowledge and text recognition without heavy data engineering, and a hybrid reinforcement learning strategy for proficiency in both short and long reasoning modes. Comprehensive experimental results in OpenCompass evaluation show that MiniCPM-V 4.5 surpasses widely used proprietary models such as GPT-4o-latest, and significantly larger open-source models such as Qwen2.5-VL 72B. Notably, the strong performance is achieved with remarkable efficiency. For example, on the widely adopted VideoMME benchmark, MiniCPM-V 4.5 achieves state-of-the-art performance among models under 30B size, using just 46.7\% GPU memory cost and 8.7\% inference time of Qwen2.5-VL 7B.

Zhengcong Fei, Zekang Li, Jinchao Zhang et al.

As a kind of new expression elements, Internet memes are popular and extensively used in online chatting scenarios since they manage to make dialogues vivid, moving, and interesting. However, most current dialogue researches focus on text-only dialogue tasks. In this paper, we propose a new task named as \textbf{M}eme incorporated \textbf{O}pen-domain \textbf{D}ialogue (MOD). Compared to previous dialogue tasks, MOD is much more challenging since it requires the model to understand the multimodal elements as well as the emotions behind them. To facilitate the MOD research, we construct a large-scale open-domain multimodal dialogue dataset incorporating abundant Internet memes into utterances. The dataset consists of $\sim$45K Chinese conversations with $\sim$606K utterances. Each conversation contains about $13$ utterances with about $4$ Internet memes on average and each utterance equipped with an Internet meme is annotated with the corresponding emotion. In addition, we present a simple and effective method, which utilizes a unified generation network to solve the MOD task. Experimental results demonstrate that our method trained on the proposed corpus is able to achieve expressive communication including texts and memes. The corpus and models have been publicly available at https://github.com/lizekang/DSTC10-MOD.

Jingyu Gong, Jiachen Xu, Xin Tan et al.

Boundary information plays a significant role in 2D image segmentation, while usually being ignored in 3D point cloud segmentation where ambiguous features might be generated in feature extraction, leading to misclassification in the transition area between two objects. In this paper, firstly, we propose a Boundary Prediction Module (BPM) to predict boundary points. Based on the predicted boundary, a boundary-aware Geometric Encoding Module (GEM) is designed to encode geometric information and aggregate features with discrimination in a neighborhood, so that the local features belonging to different categories will not be polluted by each other. To provide extra geometric information for boundary-aware GEM, we also propose a light-weight Geometric Convolution Operation (GCO), making the extracted features more distinguishing. Built upon the boundary-aware GEM, we build our network and test it on benchmarks like ScanNet v2, S3DIS. Results show our methods can significantly improve the baseline and achieve state-of-the-art performance. Code is available at https://github.com/JchenXu/BoundaryAwareGEM.

Hang Yin, Haoyu Wei, Xiuwei Xu et al.

In this paper, we propose a training-free framework for vision-and-language navigation (VLN). Existing zero-shot VLN methods are mainly designed for discrete environments or involve unsupervised training in continuous simulator environments, which makes it challenging to generalize and deploy them in real-world scenarios. To achieve a training-free framework in continuous environments, our framework formulates navigation guidance as graph constraint optimization by decomposing instructions into explicit spatial constraints. The constraint-driven paradigm decodes spatial semantics through constraint solving, enabling zero-shot adaptation to unseen environments. Specifically, we construct a spatial constraint library covering all types of spatial relationship mentioned in VLN instructions. The human instruction is decomposed into a directed acyclic graph, with waypoint nodes, object nodes and edges, which are used as queries to retrieve the library to build the graph constraints. The graph constraint optimization is solved by the constraint solver to determine the positions of waypoints, obtaining the robot's navigation path and final goal. To handle cases of no solution or multiple solutions, we construct a navigation tree and the backtracking mechanism. Extensive experiments on standard benchmarks demonstrate significant improvements in success rate and navigation efficiency compared to state-of-the-art zero-shot VLN methods. We further conduct real-world experiments to show that our framework can effectively generalize to new environments and instruction sets, paving the way for a more robust and autonomous navigation framework.

Changyuan Wang, Ziwei Wang, Xiuwei Xu et al.

In this paper, we propose an accurate data-free post-training quantization framework of diffusion models (ADP-DM) for efficient image generation. Conventional data-free quantization methods learn shared quantization functions for tensor discretization regardless of the generation timesteps, while the activation distribution differs significantly across various timesteps. The calibration images are acquired in random timesteps which fail to provide sufficient information for generalizable quantization function learning. Both issues cause sizable quantization errors with obvious image generation performance degradation. On the contrary, we design group-wise quantization functions for activation discretization in different timesteps and sample the optimal timestep for informative calibration image generation, so that our quantized diffusion model can reduce the discretization errors with negligible computational overhead. Specifically, we partition the timesteps according to the importance weights of quantization functions in different groups, which are optimized by differentiable search algorithms. We also select the optimal timestep for calibration image generation by structural risk minimizing principle in order to enhance the generalization ability in the deployment of quantized diffusion model. Extensive experimental results show that our method outperforms the state-of-the-art post-training quantization of diffusion model by a sizable margin with similar computational cost.

Zhongjin Luo, Jie Zhou, Heming Zhu et al.

Head shapes play an important role in 3D character design. In this work, we propose SimpModeling, a novel sketch-based system for helping users, especially amateur users, easily model 3D animalmorphic heads - a prevalent kind of heads in character design. Although sketching provides an easy way to depict desired shapes, it is challenging to infer dense geometric information from sparse line drawings. Recently, deepnet-based approaches have been taken to address this challenge and try to produce rich geometric details from very few strokes. However, while such methods reduce users' workload, they would cause less controllability of target shapes. This is mainly due to the uncertainty of the neural prediction. Our system tackles this issue and provides good controllability from three aspects: 1) we separate coarse shape design and geometric detail specification into two stages and respectively provide different sketching means; 2) in coarse shape designing, sketches are used for both shape inference and geometric constraints to determine global geometry, and in geometric detail crafting, sketches are used for carving surface details; 3) in both stages, we use the advanced implicit-based shape inference methods, which have strong ability to handle the domain gap between freehand sketches and synthetic ones used for training. Experimental results confirm the effectiveness of our method and the usability of our interactive system. We also contribute to a dataset of high-quality 3D animal heads, which are manually created by artists.

Xiaoyang Zheng, Xin Tan, Jie Zhou et al.

Salient object detection aims at detecting the most visually distinct objects and producing the corresponding masks. As the cost of pixel-level annotations is high, image tags are usually used as weak supervisions. However, an image tag can only be used to annotate one class of objects. In this paper, we introduce saliency subitizing as the weak supervision since it is class-agnostic. This allows the supervision to be aligned with the property of saliency detection, where the salient objects of an image could be from more than one class. To this end, we propose a model with two modules, Saliency Subitizing Module (SSM) and Saliency Updating Module (SUM). While SSM learns to generate the initial saliency masks using the subitizing information, without the need for any unsupervised methods or some random seeds, SUM helps iteratively refine the generated saliency masks. We conduct extensive experiments on five benchmark datasets. The experimental results show that our method outperforms other weakly-supervised methods and even performs comparably to some fully-supervised methods.

Jie Zhou, Will Wei Sun, Jingfei Zhang et al.

In modern data science, dynamic tensor data is prevailing in numerous applications. An important task is to characterize the relationship between such dynamic tensor and external covariates. However, the tensor data is often only partially observed, rendering many existing methods inapplicable. In this article, we develop a regression model with partially observed dynamic tensor as the response and external covariates as the predictor. We introduce the low-rank, sparsity and fusion structures on the regression coefficient tensor, and consider a loss function projected over the observed entries. We develop an efficient non-convex alternating updating algorithm, and derive the finite-sample error bound of the actual estimator from each step of our optimization algorithm. Unobserved entries in tensor response have imposed serious challenges. As a result, our proposal differs considerably in terms of estimation algorithm, regularity conditions, as well as theoretical properties, compared to the existing tensor completion or tensor response regression solutions. We illustrate the efficacy of our proposed method using simulations, and two real applications, a neuroimaging dementia study and a digital advertising study.

Zekang Li, Zongjia Li, Jinchao Zhang et al.

Audio-Visual Scene-Aware Dialog (AVSD) is a task to generate responses when chatting about a given video, which is organized as a track of the 8th Dialog System Technology Challenge (DSTC8). To solve the task, we propose a universal multimodal transformer and introduce the multi-task learning method to learn joint representations among different modalities as well as generate informative and fluent responses. Our method extends the natural language generation pre-trained model to multimodal dialogue generation task. Our system achieves the best performance in both objective and subjective evaluations in the challenge.

Jiachen Xu, Jingyu Gong, Jie Zhou et al.

Besides local features, global information plays an essential role in semantic segmentation, while recent works usually fail to explicitly extract the meaningful global information and make full use of it. In this paper, we propose a SceneEncoder module to impose a scene-aware guidance to enhance the effect of global information. The module predicts a scene descriptor, which learns to represent the categories of objects existing in the scene and directly guides the point-level semantic segmentation through filtering out categories not belonging to this scene. Additionally, to alleviate segmentation noise in local region, we design a region similarity loss to propagate distinguishing features to their own neighboring points with the same label, leading to the enhancement of the distinguishing ability of point-wise features. We integrate our methods into several prevailing networks and conduct extensive experiments on benchmark datasets ScanNet and ShapeNet. Results show that our methods greatly improve the performance of baselines and achieve state-of-the-art performance.

Xingyi Cheng, Ruiqing Zhang, Jie Zhou et al.

Predicting traffic conditions has been recently explored as a way to relieve traffic congestion. Several pioneering approaches have been proposed based on traffic observations of the target location as well as its adjacent regions, but they obtain somewhat limited accuracy due to a lack of mining road topology. To address the effect attenuation problem, we suggest taking into account the traffic of surrounding locations(wider than the adjacent range). We propose an end-to-end framework called DeepTransport, in which Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) are utilized to obtain spatial-temporal traffic information within a transport network topology. In addition, an attention mechanism is introduced to align spatial and temporal information. Moreover, we constructed and released a real-world large traffic condition dataset with a 5-minute resolution. Our experiments on this dataset demonstrate our method captures the complex relationship in the temporal and spatial domains. It significantly outperforms traditional statistical methods and a state-of-the-art deep learning method.

Yunlong Bian, Chuang Gan, Xiao Liu et al.

This paper describes our solution for the video recognition task of ActivityNet Kinetics challenge that ranked the 1st place. Most of existing state-of-the-art video recognition approaches are in favor of an end-to-end pipeline. One exception is the framework of DevNet. The merit of DevNet is that they first use the video data to learn a network (i.e. fine-tuning or training from scratch). Instead of directly using the end-to-end classification scores (e.g. softmax scores), they extract the features from the learned network and then fed them into the off-the-shelf machine learning models to conduct video classification. However, the effectiveness of this line work has long-term been ignored and underestimated. In this submission, we extensively use this strategy. Particularly, we investigate four temporal modeling approaches using the learned features: Multi-group Shifting Attention Network, Temporal Xception Network, Multi-stream sequence Model and Fast-Forward Sequence Model. Experiment results on the challenging Kinetics dataset demonstrate that our proposed temporal modeling approaches can significantly improve existing approaches in the large-scale video recognition tasks. Most remarkably, our best single Multi-group Shifting Attention Network can achieve 77.7% in term of top-1 accuracy and 93.2% in term of top-5 accuracy on the validation set.

Fu Li, Chuang Gan, Xiao Liu et al.

This paper describes our solution for the video recognition task of the Google Cloud and YouTube-8M Video Understanding Challenge that ranked the 3rd place. Because the challenge provides pre-extracted visual and audio features instead of the raw videos, we mainly investigate various temporal modeling approaches to aggregate the frame-level features for multi-label video recognition. Our system contains three major components: two-stream sequence model, fast-forward sequence model and temporal residual neural networks. Experiment results on the challenging Youtube-8M dataset demonstrate that our proposed temporal modeling approaches can significantly improve existing temporal modeling approaches in the large-scale video recognition tasks. To be noted, our fast-forward LSTM with a depth of 7 layers achieves 82.75% in term of GAP@20 on the Kaggle Public test set.

Haoyuan Gao, Junhua Mao, Jie Zhou et al.

In this paper, we present the mQA model, which is able to answer questions about the content of an image. The answer can be a sentence, a phrase or a single word. Our model contains four components: a Long Short-Term Memory (LSTM) to extract the question representation, a Convolutional Neural Network (CNN) to extract the visual representation, an LSTM for storing the linguistic context in an answer, and a fusing component to combine the information from the first three components and generate the answer. We construct a Freestyle Multilingual Image Question Answering (FM-IQA) dataset to train and evaluate our mQA model. It contains over 150,000 images and 310,000 freestyle Chinese question-answer pairs and their English translations. The quality of the generated answers of our mQA model on this dataset is evaluated by human judges through a Turing Test. Specifically, we mix the answers provided by humans and our model. The human judges need to distinguish our model from the human. They will also provide a score (i.e. 0, 1, 2, the larger the better) indicating the quality of the answer. We propose strategies to monitor the quality of this evaluation process. The experiments show that in 64.7% of cases, the human judges cannot distinguish our model from humans. The average score is 1.454 (1.918 for human). The details of this work, including the FM-IQA dataset, can be found on the project page: http://idl.baidu.com/FM-IQA.html