Visual Program Distillation: Distilling Tools and Programmatic Reasoning into Vision-Language Models
Solving complex visual tasks such as "Who invented the musical instrument on the right?" involves a composition of skills: understanding space, recognizing instruments, and also retrieving prior knowledge. Recent work shows promise by decomposing such tasks using a large language model (LLM) into an executable program that invokes specialized vision models. However, generated programs are error-prone: they omit necessary steps, include spurious ones, and are unable to recover when the specialized models give incorrect outputs. Moreover, they require loading multiple models, incurring high latency and computation costs. We propose Visual Program Distillation (VPD), an instruction tuning framework that produces a vision-language model (VLM) capable of solving complex visual tasks with a single forward pass. VPD distills the reasoning ability of LLMs by using them to sample multiple candidate programs, which are then executed and verified to identify a correct one. It translates each correct program into a language description of the reasoning steps, which are then distilled into a VLM. Extensive experiments show that VPD improves the VLM's ability to count, understand spatial relations, and reason compositionally. Our VPD-trained PaLI-X outperforms all prior VLMs, achieving state-of-the-art performance across complex vision tasks, including MMBench, OK-VQA, A-OKVQA, TallyQA, POPE, and Hateful Memes. An evaluation with human annotators also confirms that VPD improves model response factuality and consistency. Finally, experiments on content moderation demonstrate that VPD is also helpful for adaptation to real-world applications with limited data.
相關內容
ACM/IEEE第23屆模型驅動工程語言和系統國際會議，是模型驅動軟件和系統工程的首要會議系列，由ACM-SIGSOFT和IEEE-TCSE支持組織。自1998年以來，模型涵蓋了建模的各個方面，從語言和方法到工具和應用程序。模特的參加者來自不同的背景，包括研究人員、學者、工程師和工業專業人士。MODELS 2019是一個論壇，參與者可以圍繞建模和模型驅動的軟件和系統交流前沿研究成果和創新實踐經驗。今年的版本將為建模社區提供進一步推進建模基礎的機會，并在網絡物理系統、嵌入式系統、社會技術系統、云計算、大數據、機器學習、安全、開源等新興領域提出建模的創新應用以及可持續性。
官網鏈接： ·
監督
·
Networking
·
AIM
·
INFORMS
·
Weakly supervised audio-visual video parsing (AVVP) methods aim to detect audible-only, visible-only, and audible-visible events using only video-level labels. Existing approaches tackle this by leveraging unimodal and cross-modal contexts. However, we argue that while cross-modal learning is beneficial for detecting audible-visible events, in the weakly supervised scenario, it negatively impacts unaligned audible or visible events by introducing irrelevant modality information. In this paper, we propose CoLeaF, a novel learning framework that optimizes the integration of cross-modal context in the embedding space such that the network explicitly learns to combine cross-modal information for audible-visible events while filtering them out for unaligned events. Additionally, as videos often involve complex class relationships, modelling them improves performance. However, this introduces extra computational costs into the network. Our framework is designed to leverage cross-class relationships during training without incurring additional computations at inference. Furthermore, we propose new metrics to better evaluate a method's capabilities in performing AVVP. Our extensive experiments demonstrate that CoLeaF significantly improves the state-of-the-art results by an average of 1.9% and 2.4% F-score on the LLP and UnAV-100 datasets, respectively.
When editing a video, a piece of attractive background music is indispensable. However, video background music generation tasks face several challenges, for example, the lack of suitable training datasets, and the difficulties in flexibly controlling the music generation process and sequentially aligning the video and music. In this work, we first propose a high-quality music-video dataset BGM909 with detailed annotation and shot detection to provide multi-modal information about the video and music. We then present evaluation metrics to assess music quality, including music diversity and alignment between music and video with retrieval precision metrics. Finally, we propose the Diff-BGM framework to automatically generate the background music for a given video, which uses different signals to control different aspects of the music during the generation process, i.e., uses dynamic video features to control music rhythm and semantic features to control the melody and atmosphere. We propose to align the video and music sequentially by introducing a segment-aware cross-attention layer. Experiments verify the effectiveness of our proposed method. The code and models are available at //github.com/sizhelee/Diff-BGM.
A song is a combination of singing voice and accompaniment. However, existing works focus on singing voice synthesis and music generation independently. Little attention was paid to explore song synthesis. In this work, we propose a novel task called text-to-song synthesis which incorporating both vocals and accompaniments generation. We develop Melodist, a two-stage text-to-song method that consists of singing voice synthesis (SVS) and vocal-to-accompaniment (V2A) synthesis. Melodist leverages tri-tower contrastive pretraining to learn more effective text representation for controllable V2A synthesis. A Chinese song dataset mined from a music website is built up to alleviate data scarcity for our research. The evaluation results on our dataset demonstrate that Melodist can synthesize songs with comparable quality and style consistency. Audio samples can be found in //text2songMelodist.github.io/Sample/.
This work addresses the inherited limitations in the current state-of-the-art 3D multi-object tracking (MOT) methods that follow the tracking-by-detection paradigm, notably trajectory estimation drift for long-occluded objects in LiDAR point cloud streams acquired by autonomous cars. In addition, the absence of adequate track legitimacy verification results in ghost track accumulation. To tackle these issues, we introduce a two-fold innovation. Firstly, we propose refinement in Kalman filter that enhances trajectory drift noise mitigation, resulting in more robust state estimation for occluded objects. Secondly, we propose a novel online track validity mechanism to distinguish between legitimate and ghost tracks combined with a multi-stage observational gating process for incoming observations. This mechanism substantially reduces ghost tracks by up to 80\% and improves HOTA by 7\%. Accordingly, we propose an online 3D MOT framework, RobMOT, that demonstrates superior performance over the top-performing state-of-the-art methods, including deep learning approaches, across various detectors with up to 3.28\% margin in MOTA and 2.36\% in HOTA. RobMOT excels under challenging conditions, such as prolonged occlusions and the tracking of distant objects, with up to 59\% enhancement in processing latency.
The task of generating dance from music is crucial, yet current methods, which mainly produce joint sequences, lead to outputs that lack intuitiveness and complicate data collection due to the necessity for precise joint annotations. We introduce a Dance Any Beat Diffusion model, namely DabFusion, that employs music as a conditional input to directly create dance videos from still images, utilizing conditional image-to-video generation principles. This approach pioneers the use of music as a conditioning factor in image-to-video synthesis. Our method unfolds in two stages: training an auto-encoder to predict latent optical flow between reference and driving frames, eliminating the need for joint annotation, and training a U-Net-based diffusion model to produce these latent optical flows guided by music rhythm encoded by CLAP. Although capable of producing high-quality dance videos, the baseline model struggles with rhythm alignment. We enhance the model by adding beat information, improving synchronization. We introduce a 2D motion-music alignment score (2D-MM Align) for quantitative assessment. Evaluated on the AIST++ dataset, our enhanced model shows marked improvements in 2D-MM Align score and established metrics. Video results can be found on our project page: //DabFusion.github.io.
We propose a new graph convolutional block, called MusGConv, specifically designed for the efficient processing of musical score data and motivated by general perceptual principles. It focuses on two fundamental dimensions of music, pitch and rhythm, and considers both relative and absolute representations of these components. We evaluate our approach on four different musical understanding problems: monophonic voice separation, harmonic analysis, cadence detection, and composer identification which, in abstract terms, translate to different graph learning problems, namely, node classification, link prediction, and graph classification. Our experiments demonstrate that MusGConv improves the performance on three of the aforementioned tasks while being conceptually very simple and efficient. We interpret this as evidence that it is beneficial to include perception-informed processing of fundamental musical concepts when developing graph network applications on musical score data.
With the rapid advancement of generative AI, multimodal deepfakes, which manipulate both audio and visual modalities, have drawn increasing public concern. Currently, deepfake detection has emerged as a crucial strategy in countering these growing threats. However, as a key factor in training and validating deepfake detectors, most existing deepfake datasets primarily focus on the visual modal, and the few that are multimodal employ outdated techniques, and their audio content is limited to a single language, thereby failing to represent the cutting-edge advancements and globalization trends in current deepfake technologies. To address this gap, we propose a novel, multilingual, and multimodal deepfake dataset: PolyGlotFake. It includes content in seven languages, created using a variety of cutting-edge and popular Text-to-Speech, voice cloning, and lip-sync technologies. We conduct comprehensive experiments using state-of-the-art detection methods on PolyGlotFake dataset. These experiments demonstrate the dataset's significant challenges and its practical value in advancing research into multimodal deepfake detection.
Singing Accompaniment Generation (SAG), which generates instrumental music to accompany input vocals, is crucial to developing human-AI symbiotic art creation systems. The state-of-the-art method, SingSong, utilizes a multi-stage autoregressive (AR) model for SAG, however, this method is extremely slow as it generates semantic and acoustic tokens recursively, and this makes it impossible for real-time applications. In this paper, we aim to develop a Fast SAG method that can create high-quality and coherent accompaniments. A non-AR diffusion-based framework is developed, which by carefully designing the conditions inferred from the vocal signals, generates the Mel spectrogram of the target accompaniment directly. With diffusion and Mel spectrogram modeling, the proposed method significantly simplifies the AR token-based SingSong framework, and largely accelerates the generation. We also design semantic projection, prior projection blocks as well as a set of loss functions, to ensure the generated accompaniment has semantic and rhythm coherence with the vocal signal. By intensive experimental studies, we demonstrate that the proposed method can generate better samples than SingSong, and accelerate the generation by at least 30 times. Audio samples and code are available at //fastsag.github.io/.
Data documents play a central role in recording, presenting, and disseminating data. Despite the proliferation of applications and systems designed to support the analysis, visualization, and communication of data, writing data documents remains a laborious process, requiring a constant back-and-forth between data processing and writing tools. Interviews with eight professionals revealed that their workflows contained numerous tedious, repetitive, and error-prone operations. The key issue that we identified is the lack of persistent connection between text and data. Thus, we developed CrossData, a prototype that treats text-data connections as persistent, interactive, first-class objects. By automatically identifying, establishing, and leveraging text-data connections, CrossData enables rich interactions to assist in the authoring of data documents. An expert evaluation with eight users demonstrated the usefulness of CrossData, showing that it not only reduced the manual effort in writing data documents but also opened new possibilities to bridge the gap between data exploration and writing.
We construct targeted audio adversarial examples on automatic speech recognition. Given any audio waveform, we can produce another that is over 99.9% similar, but transcribes as any phrase we choose (at a rate of up to 50 characters per second). We apply our iterative optimization-based attack to Mozilla's implementation DeepSpeech end-to-end, and show it has a 100% success rate. The feasibility of this attack introduce a new domain to study adversarial examples.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191