Segmenting long videos into chapters enables users to quickly navigate to the information of their interest. This important topic has been understudied due to the lack of publicly released datasets. To address this issue, we present VidChapters-7M, a dataset of 817K user-chaptered videos including 7M chapters in total. VidChapters-7M is automatically created from videos online in a scalable manner by scraping user-annotated chapters and hence without any additional manual annotation. We introduce the following three tasks based on this data. First, the video chapter generation task consists of temporally segmenting the video and generating a chapter title for each segment. To further dissect the problem, we also define two variants of this task: video chapter generation given ground-truth boundaries, which requires generating a chapter title given an annotated video segment, and video chapter grounding, which requires temporally localizing a chapter given its annotated title. We benchmark both simple baselines and state-of-the-art video-language models for these three tasks. We also show that pretraining on VidChapters-7M transfers well to dense video captioning tasks in both zero-shot and finetuning settings, largely improving the state of the art on the YouCook2 and ViTT benchmarks. Finally, our experiments reveal that downstream performance scales well with the size of the pretraining dataset. Our dataset, code, and models are publicly available at //antoyang.github.io/vidchapters.html.
相關內容
Latent Consistency Models (LCMs) have achieved impressive performance in accelerating text-to-image generative tasks, producing high-quality images with minimal inference steps. LCMs are distilled from pre-trained latent diffusion models (LDMs), requiring only ~32 A100 GPU training hours. This report further extends LCMs' potential in two aspects: First, by applying LoRA distillation to Stable-Diffusion models including SD-V1.5, SSD-1B, and SDXL, we have expanded LCM's scope to larger models with significantly less memory consumption, achieving superior image generation quality. Second, we identify the LoRA parameters obtained through LCM distillation as a universal Stable-Diffusion acceleration module, named LCM-LoRA. LCM-LoRA can be directly plugged into various Stable-Diffusion fine-tuned models or LoRAs without training, thus representing a universally applicable accelerator for diverse image generation tasks. Compared with previous numerical PF-ODE solvers such as DDIM, DPM-Solver, LCM-LoRA can be viewed as a plug-in neural PF-ODE solver that possesses strong generalization abilities. Project page: //github.com/luosiallen/latent-consistency-model.
Synthesizing photorealistic 4D human head avatars from videos is essential for VR/AR, telepresence, and video game applications. Although existing Neural Radiance Fields (NeRF)-based methods achieve high-fidelity results, the computational expense limits their use in real-time applications. To overcome this limitation, we introduce BakedAvatar, a novel representation for real-time neural head avatar synthesis, deployable in a standard polygon rasterization pipeline. Our approach extracts deformable multi-layer meshes from learned isosurfaces of the head and computes expression-, pose-, and view-dependent appearances that can be baked into static textures for efficient rasterization. We thus propose a three-stage pipeline for neural head avatar synthesis, which includes learning continuous deformation, manifold, and radiance fields, extracting layered meshes and textures, and fine-tuning texture details with differential rasterization. Experimental results demonstrate that our representation generates synthesis results of comparable quality to other state-of-the-art methods while significantly reducing the inference time required. We further showcase various head avatar synthesis results from monocular videos, including view synthesis, face reenactment, expression editing, and pose editing, all at interactive frame rates.
Recent advances such as LLaVA and Mini-GPT4 have successfully integrated visual information into LLMs, yielding inspiring outcomes and giving rise to a new generation of multi-modal LLMs, or MLLMs. Nevertheless, these methods struggle with hallucinations and the mutual interference between tasks. To tackle these problems, we propose an efficient and accurate approach to adapt to downstream tasks by utilizing LLM as a bridge to connect multiple expert models, namely u-LLaVA. Firstly, we incorporate the modality alignment module and multi-task modules into LLM. Then, we reorganize or rebuild multi-type public datasets to enable efficient modality alignment and instruction following. Finally, task-specific information is extracted from the trained LLM and provided to different modules for solving downstream tasks. The overall framework is simple, effective, and achieves state-of-the-art performance across multiple benchmarks. We also release our model, the generated data, and the code base publicly available.
In recent years, audio-driven 3D facial animation has gained significant attention, particularly in applications such as virtual reality, gaming, and video conferencing. However, accurately modeling the intricate and subtle dynamics of facial expressions remains a challenge. Most existing studies approach the facial animation task as a single regression problem, which often fail to capture the intrinsic inter-modal relationship between speech signals and 3D facial animation and overlook their inherent consistency. Moreover, due to the limited availability of 3D-audio-visual datasets, approaches learning with small-size samples have poor generalizability that decreases the performance. To address these issues, in this study, we propose a cross-modal dual-learning framework, termed DualTalker, aiming at improving data usage efficiency as well as relating cross-modal dependencies. The framework is trained jointly with the primary task (audio-driven facial animation) and its dual task (lip reading) and shares common audio/motion encoder components. Our joint training framework facilitates more efficient data usage by leveraging information from both tasks and explicitly capitalizing on the complementary relationship between facial motion and audio to improve performance. Furthermore, we introduce an auxiliary cross-modal consistency loss to mitigate the potential over-smoothing underlying the cross-modal complementary representations, enhancing the mapping of subtle facial expression dynamics. Through extensive experiments and a perceptual user study conducted on the VOCA and BIWI datasets, we demonstrate that our approach outperforms current state-of-the-art methods both qualitatively and quantitatively. We have made our code and video demonstrations available at //github.com/sabrina-su/iadf.git.
In the last few years, deep neural networks opened the doors for big advances in novel view synthesis. Many of these approaches are based on a (coarse) proxy geometry obtained by structure from motion algorithms. Small deficiencies in this proxy can be fixed by neural rendering, but larger holes or missing parts, as they commonly appear for thin structures or for glossy regions, still lead to distracting artifacts and temporal instability. In this paper, we present a novel neural-rendering-based approach to detect and fix such deficiencies. As a proxy, we use a point cloud, which allows us to easily remove outlier geometry and to fill in missing geometry without complicated topological operations. Keys to our approach are (i) a differentiable, blending point-based renderer that can blend out redundant points, as well as (ii) the concept of Visual Error Tomography (VET), which allows us to lift 2D error maps to identify 3D-regions lacking geometry and to spawn novel points accordingly. Furthermore, (iii) by adding points as nested environment maps, our approach allows us to generate high-quality renderings of the surroundings in the same pipeline. In our results, we show that our approach can improve the quality of a point cloud obtained by structure from motion and thus increase novel view synthesis quality significantly. In contrast to point growing techniques, the approach can also fix large-scale holes and missing thin structures effectively. Rendering quality outperforms state-of-the-art methods and temporal stability is significantly improved, while rendering is possible at real-time frame rates.
Advances in ML have motivated the design of video analytics systems that allow for structured queries over video datasets. However, existing systems limit query expressivity, require users to specify an ML model per predicate, rely on complex optimizations that trade off accuracy for performance, and return large amounts of redundant and low-quality results. This paper focuses on the recently developed Vision-Language Models (VLMs) that allow users to query images using natural language like "cars during daytime at traffic intersections." Through an in-depth analysis, we show VLMs address three limitations of current video analytics systems: general expressivity, a single general purpose model to query many predicates, and are both simple and fast. However, VLMs still return large numbers of redundant and low-quality results that can overwhelm and burden users. In addition, VLMs often require manual prompt engineering to improve result relevance. We present Zelda: a video analytics system that uses VLMs to return both relevant and semantically diverse results for top-K queries on large video datasets. Zelda prompts the VLM with the user's query in natural language. Zelda then automatically adds discriminator and synonym terms to boost accuracy, and terms to identify low-quality frames. To improve result diversity, Zelda uses semantic-rich VLM embeddings in an algorithm that prunes similar frames while considering their relevance to the query and the number of top-K results requested. We evaluate Zelda across five datasets and 19 queries and quantitatively show it achieves higher mean average precision (up to 1.15x) and improves average pairwise similarity (up to 1.16x) compared to using VLMs out-of-the-box. We also compare Zelda to a state-of-the-art video analytics engine and show that Zelda retrieves results 7.5x (up to 10.4x) faster for the same accuracy and frame diversity.
Existing video compression (VC) methods primarily aim to reduce the spatial and temporal redundancies between consecutive frames in a video while preserving its quality. In this regard, previous works have achieved remarkable results on videos acquired under specific settings such as instant (known) exposure time and shutter speed which often result in sharp videos. However, when these methods are evaluated on videos captured under different temporal priors, which lead to degradations like motion blur and low frame rate, they fail to maintain the quality of the contents. In this work, we tackle the VC problem in a general scenario where a given video can be blurry due to predefined camera settings or dynamics in the scene. By exploiting the natural trade-off between visual enhancement and data compression, we formulate VC as a min-max optimization problem and propose an effective framework and training strategy to tackle the problem. Extensive experimental results on several benchmark datasets confirm the effectiveness of our method compared to several state-of-the-art VC approaches.
The proliferation of deepfake videos, synthetic media produced through advanced Artificial Intelligence techniques has raised significant concerns across various sectors, encompassing realms such as politics, entertainment, and security. In response, this research introduces an innovative and streamlined model designed to classify deepfake videos generated by five distinct encoders adeptly. Our approach not only achieves state of the art performance but also optimizes computational resources. At its core, our solution employs part of a VGG19bn as a backbone to efficiently extract features, a strategy proven effective in image-related tasks. We integrate a Capsule Network coupled with a Spatial Temporal attention mechanism to bolster the model's classification capabilities while conserving resources. This combination captures intricate hierarchies among features, facilitating robust identification of deepfake attributes. Delving into the intricacies of our innovation, we introduce an existing video level fusion technique that artfully capitalizes on temporal attention mechanisms. This mechanism serves to handle concatenated feature vectors, capitalizing on the intrinsic temporal dependencies embedded within deepfake videos. By aggregating insights across frames, our model gains a holistic comprehension of video content, resulting in more precise predictions. Experimental results on an extensive benchmark dataset of deepfake videos called DFDM showcase the efficacy of our proposed method. Notably, our approach achieves up to a 4 percent improvement in accurately categorizing deepfake videos compared to baseline models, all while demanding fewer computational resources.
Dataflow architectures are growing in popularity due to their potential to mitigate the challenges posed by the memory wall inherent to the Von Neumann architecture. At the same time, high-level synthesis (HLS) has demonstrated its efficacy as a design methodology for generating efficient dataflow architectures within a short development cycle. However, existing HLS tools rely on developers to explore the vast dataflow design space, ultimately leading to suboptimal designs. This phenomenon is especially concerning as the size of the HLS design grows. To tackle these challenges, we introduce HIDA, a new scalable and hierarchical HLS framework that can systematically convert an algorithmic description into a dataflow implementation on hardware. We first propose a collection of efficient and versatile dataflow representations for modeling the hierarchical dataflow structure. Capitalizing on these representations, we develop an automated optimizer that decomposes the dataflow optimization problem into multiple levels based on the inherent dataflow hierarchy. Using FPGAs as an evaluation platform, working with a set of neural networks modeled in PyTorch, HIDA achieves up to 8.54$\times$ higher throughput compared to the state-of-the-art (SOTA) HLS optimization tool. Furthermore, despite being fully automated and able to handle various applications, HIDA achieves 1.29$\times$ higher throughput over the SOTA RTL-based neural network accelerators on an FPGA.
Denoising diffusion models represent a recent emerging topic in computer vision, demonstrating remarkable results in the area of generative modeling. A diffusion model is a deep generative model that is based on two stages, a forward diffusion stage and a reverse diffusion stage. In the forward diffusion stage, the input data is gradually perturbed over several steps by adding Gaussian noise. In the reverse stage, a model is tasked at recovering the original input data by learning to gradually reverse the diffusion process, step by step. Diffusion models are widely appreciated for the quality and diversity of the generated samples, despite their known computational burdens, i.e. low speeds due to the high number of steps involved during sampling. In this survey, we provide a comprehensive review of articles on denoising diffusion models applied in vision, comprising both theoretical and practical contributions in the field. First, we identify and present three generic diffusion modeling frameworks, which are based on denoising diffusion probabilistic models, noise conditioned score networks, and stochastic differential equations. We further discuss the relations between diffusion models and other deep generative models, including variational auto-encoders, generative adversarial networks, energy-based models, autoregressive models and normalizing flows. Then, we introduce a multi-perspective categorization of diffusion models applied in computer vision. Finally, we illustrate the current limitations of diffusion models and envision some interesting directions for future research.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191