Real-world multi-modal problems are rarely solved by a single machine learning model, and often require multi-step computational plans that involve stitching several models. Tool-augmented LLMs hold tremendous promise for automating the generation of such computational plans. However, the lack of standardized benchmarks for evaluating LLMs as planners for multi-step multi-modal tasks has prevented a systematic study of planner design decisions. Should LLMs generate a full plan in a single shot or step-by-step? Should they invoke tools directly with Python code or through structured data formats like JSON? Does feedback improve planning? To answer these questions and more, we introduce m&m’s: a benchmark containing 4K+ multi-step multi-modal tasks involving 33 tools that include multi-modal models, (free) public APIs, and image processing modules. For each of these task queries, we provide automatically generated plans using this realistic toolset. We further provide a high-quality subset of 1565 task plans that are human-verified and correctly executable. With m&m’s, we evaluate 6 popular LLMs with 2 planning strategies (multi-step vs. step-by-step planning), 2 plan formats (JSON vs. code), and 3 types of feedback (parsing/verification/execution). Finally, we summarize takeaways from our extensive experiments and provide practical recommendations for designing planners for m&m’s tasks.
论文检索
输入标题、作者或关键词，从 3,655 篇学术成果中精准定位
Cross-view geo-localization serves as a viable alternative to providing geographical location information when GPS signals are unstable or unavailable by matching ground images with geo-tagged aerial image databases. While significant progress has been made on some common benchmarks like CVUSA and CVACT, there remains a lack of comprehensive consideration for robustness against real-world environmental challenges such as adverse weather or sensor noise. This deficiency poses a significant challenge for deploying this technology in safety-critical domains like autonomous driving and robot navigation. To the best of our knowledge, there is currently no specialized benchmark for evaluating the robustness of cross-view geo-localization. To comprehensively and fairly evaluate the robustness of cross-view geo-localization models in real-world scenarios, we introduce 16 common types of data corruption. By synthesizing these corruptions on public datasets, we establish two fine-grained corruption robustness benchmarks (CVUSA-C and CVACTval-C) and three comprehensive corruption robustness benchmarks (CVUSA-C-ALL, CVACTval-C-ALL, and CVACT_test-C-ALL), covering approximately 1.5 million corrupted images. Subsequently, we conduct large-scale experiments on various cross-view geo-localization models to evaluate their robustness in corrupted environments and derive novel insights. Finally, we explore two data augmentation strategies as potential solutions to enhance model robustness. Combined with the training strategies proposed, these approaches effectively enhance the robustness of multiple models. We provide benchmarks and codes in the supplementary materials to be helpful for future studies.
Predicting future trajectories for other road agents is an essential task for autonomous vehicles. Established trajectory prediction methods primarily use agent tracks generated by a detection and tracking system and HD map as inputs to a model which predicts agent trajectories. In this work, we propose a novel method that also incorporates visual input from surround-view cameras, allowing the model to utilize visual cues such as human gazes and gestures, road conditions, vehicle turn signals, etc, which are typically hidden from the model in prior trajectory prediction methods. Furthermore, we use textual descriptions generated by a Vision-Language Model (VLM) and refined by a Large Language Model (LLM) as supervision to guide the model on what to learn from the input data. Our experiments show that both the visual inputs and the textual descriptions contribute to improvements in trajectory prediction performance, and our qualitative analysis highlights how the model is able to exploit these additional inputs. Despite using these extra inputs, our method achieves a latency of 53 ms, significantly lower than that of previous single-agent prediction methods with similar performance. Lastly, in this work we create and release the nuScenes-Text dataset, which augments the established nuScenes dataset with rich textual annotations for every scene, demonstrating the positive impact of utilizing VLM on trajectory prediction.
Post-Training Quantization (PTQ) has received significant attention because it requires only a small set of calibration data to quantize a full-precision model, which is more practical in real-world applications in which full access to a large training set is not available. However, it often leads to overfitting on the small calibration dataset. Several methods have been proposed to address this issue, yet they still rely on only the calibration set for the quantization and they do not validate the quantized model due to the lack of a validation set. In this work, we propose a novel meta-learning based approach to enhance the performance of post-training quantization. Specifically, to mitigate the overfitting problem, instead of only training the quantized model using the original calibration set without any validation during the learning process as in previous PTQ works, in our approach, we both train and validate the quantized model using two different sets of images. In particular, we propose a meta-learning based approach to jointly optimize a transformation network and a quantized model through bi-level optimization. The transformation network modifies the original calibration data and the modified data will be used as the training set to learn the quantized model with the objective that the quantized model achieves a good performance on the original calibration data. Extensive experiments on the widely used ImageNet dataset with different neural network architectures demonstrate that our approach outperforms the state-of-the-art PTQ methods.
Domain Generalization (DG), designed to enhance out-of-distribution (OOD) generalization, is all about learning invariance against domain shifts utilizing sufficient supervision signals. Yet, the scarcity of such labeled data has led to the rise of unsupervised domain generalization (UDG) — a more important yet challenging task in that models are trained across diverse domains in an unsupervised manner and eventually tested on unseen domains. UDG is fast gaining attention but is still far from well-studied. To close the research gap, we propose a novel learning framework designed for UDG, termed the Disentangled Masked AutoEncoder (DisMAE), aiming to discover the disentangled representations that faithfully reveal the intrinsic features and superficial variations without access to the class label. At its core is the distillation of domain-invariant semantic features, which cannot be distinguished by domain classifier, while filtering out the domain-specific variations (for example, color schemes and texture patterns) that are unstable and redundant. Notably, DisMAE co-trains the asymmetric dual-branch architecture with semantic and lightweight variation encoders, offering dynamic data manipulation and representation level augmentation capabilities. Extensive experiments on four benchmark datasets (i.e., DomainNet, PACS, VLCS, and Colored MNIST) with both DG and UDG tasks demonstrate that DisMAE can achieve competitive OOD performance compared with the state-of-the-art DG and UDG baselines, which shed light on potential research lines for improving generalization ability with large-scale unlabeled data.
This work focuses on generating realistic, physically-based human behaviors from multi-modal inputs, which may only partially specify the desired motion. For example, the input may come from a VR controller providing arm motion and body velocity, partial key-point animation, computer vision applied to videos, or even higher-level motion goals. This requires a versatile low-level humanoid controller that can handle such sparse, under-specified guidance, seamlessly switch between skills, and recover from failures. Current approaches for learning humanoid controllers from demonstration data capture some of these characteristics, but none achieve them all. To this end, we introduce the Masked Humanoid Controller (MHC), a novel approach that applies multi-objective imitation learning on augmented and selectively masked motion demonstrations. The training methodology results in an MHC that exhibits the key capabilities of catch-up to out-of-sync input commands, combining elements from multiple motion sequences, and completing unspecified parts of motions from sparse multimodal input. We demonstrate these key capabilities for an MHC learned over a dataset of 87 diverse skills and showcase different multi-modal use cases, including integration with planning frameworks to highlight MHC’s ability to solve new user-defined tasks without any finetuning.
With the development of AIGC, images generated by AI are difficult to identify with the human eye. This article proposes a simple training architecture and a simple model. In the training process of deep forgery detection, a variety of data enhancement techniques are used, including horizontal flipping, adding Gaussian noise and random cropping to enhance the generalization ability of the model. Differentiated processing strategies are adopted for images of different sizes: center cropping is used for images smaller than 224 pixels, random cropping is used for images between 224 and 512 pixels, and images larger than 512 pixels are first scaled and then randomly cropped. Furthermore, the concept of neighbor pixel relations (NPR) is introduced, which is an effective way to capture local structural artifacts introduced by upsampling operations. As a complement to image representation, NPR significantly improves the detector's ability to identify forged images, especially when the forgery technology is unknown. By integrating the NPR module into the ResNet-50 model, it not only enhances the classifier's extraction of forged features, but also maintains the computational efficiency and versatility of the model. Finally, AIGC generated images in the ''Malanshan Cup" International Audio-Video Algorithm Competition. Determine the first place result for the track.
In recent years, the task of image-to-text generation has received considerable attention from scholars. One of its subtasks, Visual Spatial Description (VSD), focuses on a model's ability to understand spatial relationships. VSD is a novel task that emphasizes spatial semantics by generating sentences describing the spatial relationships between two objects in a given image. In this work, a VSD method based on large language model fine-tuning (LFVSD) is proposed to enhance the accuracy and robustness of visual spatial relationship descriptions. Initially, image and text features are extracted using pre-trained models, and Q-Former is employed for feature fusion. The original and fused features are then fed into FlanT5XXL. Object overlap priors are introduced, and momentum distillation is used to filter hard negative samples and generate soft labels. Finally, multiple VSD models are trained using data augmentation and long-tail data balancing techniques. Through multimodal feature fusion and fine-tuning, our approach is evaluated on the VSD2024 test set, which includes 5,855 images and their corresponding textual descriptions. The results demonstrate the effectiveness of our proposed method.
The continual advancements in Generative Artificial Intelligence have created substantial hurdles for accurate deepfake detection, leading to limitations of currently popular detection methods across content-driven video-level deepfake detection scenarios. In this paper, we present the solutions to the Video-Level Deepfake Detection task. Our empirical findings demonstrate that modeling correlations of audio-visual modalities is important for video-level deepfake detection. Therefore, we introduce the model denoted Audio-Visual Local-Global Neural Network (i.e., AV-LGNN) in which the core design is the proposed AV-LGI Module (Audio-Visual Local-Global Interaction Module). The AV-LGI Module is composed of three stages: Local Intra-Region Interaction, Global Inter-Region Interaction, and Local-Global Interaction, which can better capture detailed information at local-level and efficiently learn the fine-grained correlations of inter-modalities in video deepfake detection under lower computational overheads. We further propose an adaptive modality selection strategy to facilitate model learning. Besides, a variety of data augmentation techniques are incorporated for audio-visual branches to enhance the robustness of the AV-LGNN. The experimental results verify the effectiveness of our model.
This paper presents Team Xaiofei's innovative approach to exploring Face-Voice Association in Multilingual Environments (FAME) at ACM Multimedia 2024. We focus on the impact of different languages in face-voice matching by building upon Fusion and Orthogonal Projection (FOP), introducing four key components: a dual-branch structure, dynamic sample pair weighting, robust data augmentation, and score polarization strategy. Our dual-branch structure serves as an auxiliary mechanism to better integrate and provide more comprehensive information. We also introduce a dynamic weighting mechanism for various sample pairs to optimize learning. Data augmentation techniques are employed to enhance the model's generalization across diverse conditions. Additionally, score polarization strategy based on age and gender matching confidence clarifies and accentuates the final results. Our methods demonstrate significant effectiveness, achieving an equal error rate (EER) of 20.07 on the V2-EH dataset and 21.76 on the V1-EU dataset. Project page: https://github.com/cnzvan/Exploring-Robust-Face-Voice-Matching-in-Multilingual-Environments.
Graph-based fraud detection (GFD) has garnered increasing attention due to its effectiveness in identifying fraudsters within multimedia data such as online transactions, product reviews, or telephone voices. However, the prevalent in-distribution (ID) assumption significantly impedes the generalization of GFD approaches to out-of-distribution (OOD) scenarios, which is a pervasive challenge considering the dynamic nature of fraudulent activities. In this paper, we introduce the Heterophilic Graph Invariant Learning Framework (HGIF), a novel approach to bolster the OOD generalization of GFD. HGIF addresses two pivotal challenges: creating diverse virtual training environments and adapting to varying target distributions. Leveraging edge-aware augmentation, HGIF efficiently generates multiple virtual training environments characterized by generalized heterophily distributions, thereby facilitating robust generalization against fraud graphs with diverse heterophily degrees. Moreover, HGIF employs a shared dual-channel encoder with heterophilic graph contrastive learning, enabling the model to acquire stable high-pass and low-pass node representations during training. During the Test-time Training phase, the shared dual-channel encoder is flexibly fine-tuned to adapt to the test distribution through graph contrastive learning. Extensive experiments showcase HGIF's superior performance over existing methods in OOD generalization, setting a new benchmark for GFD in OOD scenarios.
The task of text semantic matching focuses on measuring the semantic similarity between two texts and is widely applied in search and ranking scenarios. In recent years, pre-trained foundation models based on the Transformer architecture have demonstrated powerful semantic representation capabilities. The pipeline of fine-tuning pre-trained foundation models on downstream semantic matching tasks has achieved promising results and widespread adoption. However, practical downstream scenarios often face severe challenges in terms of data quality and quantity. Ensuring high-quality and large quantities of samples is often difficult. Current research on enhancing pre-trained models for few-shot text semantic matching tasks is still not advanced enough. Therefore, this paper focuses on providing a general enhancement scheme for few-shot text semantic matching tasks. Specifically, we propose an Enhancing Transformer-based Semantic Matching method for few-shot learning through weakly contrastive pre-training, which is named as SEMFormer. First, starting from the token-level and structural-level perspectives, we design a simple and low-cost data augmentation method to construct weakly supervised samples. Then, we use the global semantic representations to construct a contrastive objective from the relation-aspect perspective. Next, we design a contrastive objective based on the alignment-aspect, aiming to achieve effective semantic matching by optimizing the bidirectional semantic awareness between texts. We conducted comprehensive experiments based on five Chinese and English datasets. The experimental results validated that our proposed weakly contrastive pre-training augmentation method significantly improves model performance. Further experiments confirmed the effectiveness of our design. The source code is available at: https://github.com/llm-ml/SEMFormer.
Latent diffusion model has demonstrated impressive efficacy in image generation and editing tasks. Recently, it has also promoted the advancement of image harmonization. However, methods involving latent diffusion model all face a common challenge: the severe image distortion introduced by the VAE component, while image harmonization is a low-level image processing task that relies on pixel-level evaluation metrics. In this paper, we propose Harmony-VAE, leveraging the input of the harmonization task itself to enhance the quality of decoded images. The input involving composite image contains the precise pixel level information, which can complement the correct foreground appearance and color information contained in denoised latents. Meanwhile, the inherent generative nature of diffusion models makes it naturally adapt to inverse image harmonization, i.e. generating synthetic composite images based on real images and foreground masks. We train an inverse harmonization diffusion model to perform data augmentation on two subsets of iHarmony4 and construct a new human harmonization dataset with prominent foreground objects. Extensive experiments demonstrate the effectiveness of our proposed Harmony-VAE and inverse harmonization model. Code and pretrained models are available at https://github.com/nicecv/DiffHarmony.
News event discovery refers to the identification and detection of news events using multimodal data on social media. Currently, most works assume that the test set consists of known events. However, in real life, the emergence of new events is more frequent, which invalidates this assumption. In this paper, we propose a Dynamic Augmentation and Entropy Optimization (DAEO) model to address the scenario of generalized news event discovery, which requires the model to not only identify known events but also distinguish various new events. Specifically, we first introduce a multimodal augmentation module, which utilizes adversarial learning to enhance the multimodal representation capability. Secondly, we design an adaptive entropy optimization strategy combined with a self-distillation method, which uses multi-view pseudo-label consistency to improve the model's performance on both known and new events. In addition, we collect a multimodal news event discovery (MNED) dataset of 161,350 samples annotated with 66 real-world events. Extensive experimental results on the MNED dataset demonstrate the effectiveness of our proposed method. Our dataset is available on https://github.com/RetrainIt/MNED.
This paper introduces HINER, a novel neural representation for compressing HSI and ensuring high-quality downstream tasks on compressed HSI. HINER fully exploits inter-spectral correlations by explicitly encoding of spectral wavelengths and achieves a compact representation of the input HSI sample through joint optimization with a learnable decoder. By additionally incorporating the Content Angle Mapper with the L1 loss, we can supervise the global and local information within each spectral band, thereby enhancing the overall reconstruction quality. For downstream classification on compressed HSI, we theoretically demonstrate the task accuracy is not only related to the classification loss but also to the reconstruction fidelity through a first-order expansion of the accuracy degradation, and accordingly adapt the reconstruction by introducing Adaptive Spectral Weighting. Owing to the monotonic mapping of HINER between wavelengths and spectral bands, we propose Implicit Spectral Interpolation for data augmentation by adding random variables to input wavelengths during classification model training. Experimental results on various HSI datasets demonstrate the superior compression performance of our HINER compared to the existing learned methods and also the traditional codecs. Our model is lightweight and computationally efficient, which maintains high accuracy for downstream classification task even on decoded HSIs at high compression ratios. Our materials will be released at https://github.com/Eric-qi/HINER.
Learning semantic-rich representations from unlabeled time series data with intricate dynamics is a notable challenge. Traditional contrastive learning techniques predominantly focus on segment-level augmentations through time slicing, a practice that, while valuable, often results in sampling bias and suboptimal performance due to the loss of global context. Furthermore, they typically disregard the vital frequency information to enrich data representations. To this end, we propose a novel self-supervised general-purpose framework called Temporal-Frequency and Contextual Consistency (TFCC). Specifically, this framework first performs two instance-level augmentation families over the entire series to capture nuanced representations alongside critical long-term dependencies. Then, TFCC advances by initiating dual cross-view forecasting tasks between the original series and its augmented counterpart in both time and frequency domains to learn robust representations. Finally, three specially designed consistency modules 'temporal, frequency, and temporal-frequency' aid in further developing discriminative representations on top of the learned robust representations. Extensive experiments on multiple benchmarks demonstrate TFCC's superiority over the state-of-the-art classification and forecasting methods and exhibit exceptional efficiency in semi-supervised and transfer learning scenarios.
Towards Robustness Prompt Tuning with Fully Test-Time Adaptation for CLIP's Zero-Shot Generalization
In the field of Vision-Language Models (VLM), the Contrastive Language-Image Pretraining (CLIP) model has yielded outstanding performance on many downstream tasks through prompt tuning. By integrating image and text representations, CLIP exhibits zero-shot generalization capabilities on unseen data. However, when new categories and distribution shifts occur, the pretrained text embeddings in CLIP may not align well with unseen images, potentially leading to a decrease in CLIP's zero-shot generalization performance. To address this issue, many existing methods use test samples to update the CLIP model during testing through a process known as Test-Time Adaptation (TTA). Previous TTA techniques, such as image augmentation, can lead to overfitting given outlying samples, while methods based on teacher-student distillation can increase memory use. Further, these methods significantly increase inference time, which is a crucial factor in the testing phase. To improve robustness, mitigate overfitting, and reduce bias toward outlying samples, we propose a novel method: Self-Text Distillation with Conjugate Pseudo-labels (SCP), designed to enhance CLIP's zero-shot generalization. SCP uses gradient information from conjugate pseudo-labels to enhance the model's robustness toward distribution shifts. It also innovates by using a fixed prompt list to distil learnable prompts from within the same model, acting as a self-regulation mechanism that minimizes overfitting. Additionally, SCP is a fully test-time adaptation method that does not require retraining. It directly improves CLIP's zero-shot generalization at test time without increasing either memory overheads or inference time. In evaluations across three zero-shot generalization scenarios, SCP surpasses existing state-of-the-art methods in performance and significantly reduces inference time.
Self-Adaptive Fine-grained Multi-modal Data Augmentation for Semi-supervised Muti-modal Coreference Resolution
PDF ↗Coreference resolution, an essential task in natural language processing, is particularly challenging in multi-modal scenarios where data comes in various forms and modalities. Despite advancements, limitations due to scarce labeled data and underleveraged unlabeled data persist. We address these issues with a self-adaptive fine-grained multi-modal data augmentation framework for semi-supervised MCR, focusing on enriching training data from labeled datasets and tapping into the untapped potential of unlabeled data. Regarding the former issue, we first leverage text coreference resolution datasets and diffusion models,to perform fine-grained text-to-image generation with aligned text entities and image bounding boxes. We then introduce a self-adaptive selection strategy, meticulously curating the augmented data to enhance the diversity and volume of the training set without compromising its quality. For the latter issue, we design a self-adaptive threshold strategy that dynamically adjusts the confidence threshold based on the model's learning status and performance, enabling effective utilization of valuable information from unlabeled data. Additionally, we incorporate a distance smoothing term, which smooths distances between positive and negative samples, enhancing discriminative power of the model?s feature representations and addressing noise and uncertainty in the unlabeled data. Our experiments on the widely-used CIN dataset show that our framework significantly outperforms state-of-the-art baselines by at least 9.57% on MUC F1 score and 4.92% on CoNLL F1 score. Remarkably, against weakly-supervised baselines, our framework achieves a staggering 22.24% enhancement in MUC F1 score. These results, underpinned by in-depth analyses, underscore the effectiveness and potential of our approach for advancing MCR tasks.
The 3D model can be estimated by regressing the pose and shape parameters from the image data of the digital model. The reconstruction of 3D cartoon characters poses a challenging task due to diverse visual representations and postural variations. This paper proposes a dual-branch structure named MagicCartoon for 3D bipedal cartoon character estimation, which models pose and shape independently through feature decoupling. Considering the correlation between category difference and shape parameters, a hybrid feature fusion technique is introduced, which integrates the global features of the original image with the corresponding local features expressed by the puzzle image, reducing the abstractness of understanding shape parameter differences. To semantically align image and geometric between feature space, a geometric-guided feedback loop is proposed in an iterative way, so that the pose of modeling results can be expressed consistently with the image. Moreover, a feature consistency loss is designed to augment the training data by incorporating the same character with different postures and the same posture of different characters. It enhances the correlation between the features extracted by the backbone network and the specific task. Experiments conducted on the 3DBiCar dataset demonstrate that MagicCartoon outperforms the state-of-the-art methods.
Bokeh is a wide-aperture optical effect that creates aesthetic blurring in photography. However, achieving this effect typically demands expensive professional equipment and expertise. To make such cinematic techniques more accessible, bokeh rendering aims to generate the desired bokeh effects from all-in-focus inputs captured by smartphones. Previous efforts in bokeh rendering primarily focus on static images. However, when extended to video inputs, these methods exhibit flicker and artifacts due to a lack of temporal consistency modeling. Meanwhile, they cannot utilize information like occluded objects from adjacent frames, which are necessary for bokeh rendering. Moreover, the difficulties of capturing all-in-focus and bokeh video pairs result in a shortage of data for training video bokeh models. To tackle these challenges, we propose the Video Bokeh Renderer (VBR), the model designed specifically for video bokeh rendering.VBR leverages implicit feature space alignment and aggregation to model temporal consistency and exploit complementary information from adjacent frames. On the data front, we introduce the first Synthetic Video Bokeh (SVB) dataset, synthesizing authentic bokeh effects using ray-tracing techniques. Furthermore, to improve the robustness of the model to inaccurate disparity maps, we employ a set of augmentation strategies to simulate corrupted disparity inputs during training. Experimental results on both synthetic and real-world data demonstrate the effectiveness of our method.