Recent advancements in 3D Gaussian Splatting (3DGS) have unlocked significant potential for modeling 3D head avatars, providing greater flexibility than mesh-based methods and more efficient rendering compared to NeRF-based approaches. Despite these advancements, the creation of controllable 3DGS-based head avatars remains time-intensive, often requiring tens of minutes to hours. To expedite this process, we here introduce the ``Gaussian D\'ej\`a-vu" framework, which first obtains a generalized model of the head avatar and then personalizes the result. The generalized model is trained on large 2D (synthetic and real) image datasets. This model provides a well-initialized 3D Gaussian head that is further refined using a monocular video to achieve the personalized head avatar. For personalizing, we propose learnable expression-aware rectification blendmaps to correct the initial 3D Gaussians, ensuring rapid convergence without the reliance on neural networks. Experiments demonstrate that the proposed method meets its objectives. It outperforms state-of-the-art 3D Gaussian head avatars in terms of photorealistic quality as well as reduces training time consumption to at least a quarter of the existing methods, producing the avatar in minutes.
相關內容
Recent advances in behavior cloning (BC), like action-chunking and diffusion, have led to impressive progress. Still, imitation alone remains insufficient for tasks requiring reliable and precise movements, such as aligning and inserting objects. Our key insight is that chunked BC policies function as trajectory planners, enabling long-horizon tasks. Conversely, as they execute action chunks open-loop, they lack the fine-grained reactivity necessary for reliable execution. Further, we find that the performance of BC policies saturates despite increasing data. Reinforcement learning (RL) is a natural way to overcome this, but it is not straightforward to apply directly to action-chunked models like diffusion policies. We present a simple yet effective method, ResiP (Residual for Precise Manipulation), that sidesteps these challenges by augmenting a frozen, chunked BC model with a fully closed-loop residual policy trained with RL. The residual policy is trained via on-policy RL, addressing distribution shifts and introducing reactivity without altering the BC trajectory planner. Evaluation on high-precision manipulation tasks demonstrates strong performance of ResiP over BC methods and direct RL fine-tuning. Videos, code, and data are available at \url{//residual-assembly.github.io}.
In Industry 4.0 systems, a considerable number of resource-constrained Industrial Internet of Things (IIoT) devices engage in frequent data interactions due to the necessity for model training, which gives rise to concerns pertaining to security and privacy. In order to address these challenges, this paper considers a digital twin (DT) and blockchain-assisted federated learning (FL) scheme. To facilitate the FL process, we initially employ fog devices with abundant computational capabilities to generate DT for resource-constrained edge devices, thereby aiding them in local training. Subsequently, we formulate an FL delay minimization problem for FL, which considers both of model transmission time and synchronization time, also incorporates cooperative jamming to ensure secure synchronization of DT. To address this non-convex optimization problem, we propose a decomposition algorithm. In particular, we introduce upper limits on the local device training delay and the effects of aggregation jamming as auxiliary variables, thereby transforming the problem into a convex optimization problem that can be decomposed for independent solution. Finally, a blockchain verification mechanism is employed to guarantee the integrity of the model uploading throughout the FL process and the identities of the participants. The final global model is obtained from the verified local and global models within the blockchain through the application of deep learning techniques. The efficacy of our proposed cooperative interference-based FL process has been verified through numerical analysis, which demonstrates that the integrated DT blockchain-assisted FL scheme significantly outperforms the benchmark schemes in terms of execution time, block optimization, and accuracy.
Recently, driven by advancements in Multimodal Large Language Models (MLLMs), Vision Language Action Models (VLAMs) are being proposed to achieve better performance in open-vocabulary scenarios for robotic manipulation tasks. Since manipulation tasks involve direct interaction with the physical world, ensuring robustness and safety during the execution of this task is always a very critical issue. In this paper, by synthesizing current safety research on MLLMs and the specific application scenarios of the manipulation task in the physical world, we comprehensively evaluate VLAMs in the face of potential physical threats. Specifically, we propose the Physical Vulnerability Evaluating Pipeline (PVEP) that can incorporate as many visual modal physical threats as possible for evaluating the physical robustness of VLAMs. The physical threats in PVEP specifically include Out-of-Distribution, Typography-based Visual Prompts, and Adversarial Patch Attacks. By comparing the performance fluctuations of VLAMs before and after being attacked, we provide generalizable Analyses of how VLAMs respond to different physical security threats. Our project page is in this link: //chaducheng.github.io/Manipulat-Facing-Threats/.
Recent self-supervised clustering-based pre-training techniques like DINO and Cribo have shown impressive results for downstream detection and segmentation tasks. However, real-world applications such as autonomous driving face challenges with imbalanced object class and size distributions and complex scene geometries. In this paper, we propose S3PT a novel scene semantics and structure guided clustering to provide more scene-consistent objectives for self-supervised training. Specifically, our contributions are threefold: First, we incorporate semantic distribution consistent clustering to encourage better representation of rare classes such as motorcycles or animals. Second, we introduce object diversity consistent spatial clustering, to handle imbalanced and diverse object sizes, ranging from large background areas to small objects such as pedestrians and traffic signs. Third, we propose a depth-guided spatial clustering to regularize learning based on geometric information of the scene, thus further refining region separation on the feature level. Our learned representations significantly improve performance in downstream semantic segmentation and 3D object detection tasks on the nuScenes, nuImages, and Cityscapes datasets and show promising domain translation properties.
The rise of Large Language Models (LLMs) has streamlined frontend interface creation through tools like Vercel's V0, yet surfaced challenges in design quality (e.g., accessibility, and usability). Current solutions, often limited by their focus, generalisability, or data dependency, fall short in addressing these complexities. Moreover, none of them examine the quality of LLM-generated UI design. In this work, we introduce DesignRepair, a novel dual-stream design guideline-aware system to examine and repair the UI design quality issues from both code aspect and rendered page aspect. We utilised the mature and popular Material Design as our knowledge base to guide this process. Specifically, we first constructed a comprehensive knowledge base encoding Google's Material Design principles into low-level component knowledge base and high-level system design knowledge base. After that, DesignRepair employs a LLM for the extraction of key components and utilizes the Playwright tool for precise page analysis, aligning these with the established knowledge bases. Finally, we integrate Retrieval-Augmented Generation with state-of-the-art LLMs like GPT-4 to holistically refine and repair frontend code through a strategic divide and conquer approach. Our extensive evaluations validated the efficacy and utility of our approach, demonstrating significant enhancements in adherence to design guidelines, accessibility, and user experience metrics.
Alongside the rapid development of Large Language Models (LLMs), there has been a notable increase in efforts to integrate LLM techniques in information retrieval (IR) and search engines (SE). Recently, an additional post-ranking stage is suggested in SE to enhance user satisfaction in practical applications. Nevertheless, research dedicated to enhancing the post-ranking stage through LLMs remains largely unexplored. In this study, we introduce a novel paradigm named Large Language Models for Post-Ranking in search engine (LLM4PR), which leverages the capabilities of LLMs to accomplish the post-ranking task in SE. Concretely, a Query-Instructed Adapter (QIA) module is designed to derive the user/item representation vectors by incorporating their heterogeneous features. A feature adaptation step is further introduced to align the semantics of user/item representations with the LLM. Finally, the LLM4PR integrates a learning to post-rank step, leveraging both a main task and an auxiliary task to fine-tune the model to adapt the post-ranking task. Experiment studies demonstrate that the proposed framework leads to significant improvements and exhibits state-of-the-art performance compared with other alternatives.
We present LRM-Zero, a Large Reconstruction Model (LRM) trained entirely on synthesized 3D data, achieving high-quality sparse-view 3D reconstruction. The core of LRM-Zero is our procedural 3D dataset, Zeroverse, which is automatically synthesized from simple primitive shapes with random texturing and augmentations (e.g., height fields, boolean differences, and wireframes). Unlike previous 3D datasets (e.g., Objaverse) which are often captured or crafted by humans to approximate real 3D data, Zeroverse completely ignores realistic global semantics but is rich in complex geometric and texture details that are locally similar to or even more intricate than real objects. We demonstrate that our LRM-Zero, trained with our fully synthesized Zeroverse, can achieve high visual quality in the reconstruction of real-world objects, competitive with models trained on Objaverse. We also analyze several critical design choices of Zeroverse that contribute to LRM-Zero's capability and training stability. Our work demonstrates that 3D reconstruction, one of the core tasks in 3D vision, can potentially be addressed without the semantics of real-world objects. The Zeroverse's procedural synthesis code and interactive visualization are available at: //desaixie.github.io/lrm-zero/.
Despite advancements in Text-to-Video (T2V) generation, producing videos with realistic motion remains challenging. Current models often yield static or minimally dynamic outputs, failing to capture complex motions described by text. This issue stems from the internal biases in text encoding, which overlooks motions, and inadequate conditioning mechanisms in T2V generation models. To address this, we propose a novel framework called DEcomposed MOtion (DEMO), which enhances motion synthesis in T2V generation by decomposing both text encoding and conditioning into content and motion components. Our method includes a content encoder for static elements and a motion encoder for temporal dynamics, alongside separate content and motion conditioning mechanisms. Crucially, we introduce text-motion and video-motion supervision to improve the model's understanding and generation of motion. Evaluations on benchmarks such as MSR-VTT, UCF-101, WebVid-10M, EvalCrafter, and VBench demonstrate DEMO's superior ability to produce videos with enhanced motion dynamics while maintaining high visual quality. Our approach significantly advances T2V generation by integrating comprehensive motion understanding directly from textual descriptions. Project page: //PR-Ryan.github.io/DEMO-project/
Recent advancements in Large Language Models (LLMs) have significantly enhanced their ability to process long contexts, yet a notable gap remains in generating long, aligned outputs. This limitation stems from a training gap where pre-training lacks effective instructions for long-text generation, and post-training data primarily consists of short query-response pairs. Current approaches, such as instruction backtranslation and behavior imitation, face challenges including data quality, copyright issues, and constraints on proprietary model usage. In this paper, we introduce an innovative iterative training framework called Self-Lengthen that leverages only the intrinsic knowledge and skills of LLMs without the need for auxiliary data or proprietary models. The framework consists of two roles: the Generator and the Extender. The Generator produces the initial response, which is then split and expanded by the Extender. This process results in a new, longer response, which is used to train both the Generator and the Extender iteratively. Through this process, the models are progressively trained to handle increasingly longer responses. Experiments on benchmarks and human evaluations show that Self-Lengthen outperforms existing methods in long-text generation, when applied to top open-source LLMs such as Qwen2 and LLaMA3. Our code is publicly available at //github.com/QwenLM/Self-Lengthen.
Vision-Language Models (VLMs) have demonstrated impressive performance across a versatile set of tasks. A key challenge in accelerating VLMs is storing and accessing the large Key-Value (KV) cache that encodes long visual contexts, such as images or videos. While existing KV cache compression methods are effective for Large Language Models (LLMs), directly migrating them to VLMs yields suboptimal accuracy and speedup. To bridge the gap, we propose VL-Cache, a novel KV cache compression recipe tailored for accelerating VLM inference. In this paper, we first investigate the unique sparsity pattern of VLM attention by distinguishing visual and text tokens in prefill and decoding phases. Based on these observations, we introduce a layer-adaptive sparsity-aware cache budget allocation method that effectively distributes the limited cache budget across different layers, further reducing KV cache size without compromising accuracy. Additionally, we develop a modality-aware token scoring policy to better evaluate the token importance. Empirical results on multiple benchmark datasets demonstrate that retaining only 10% of KV cache achieves accuracy comparable to that with full cache. In a speed benchmark, our method accelerates end-to-end latency of generating 100 tokens by up to 2.33x and speeds up decoding by up to 7.08x, while reducing the memory footprint of KV cache in GPU by 90%.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191