Deepfakes are AI-synthesized multimedia data that may be abused for spreading misinformation. Deepfake generation involves both visual and audio manipulation. To detect audio-visual deepfakes, previous studies commonly employ two relatively independent sub-models to learn audio and visual features, respectively, and fuse them subsequently for deepfake detection. However, this may underutilize the inherent correlations between audio and visual features. Moreover, utilizing two isolated feature learning sub-models can result in redundant neural layers, making the overall model inefficient and impractical for resource-constrained environments. In this work, we design a lightweight network for audio-visual deepfake detection via a single-stream multi-modal learning framework. Specifically, we introduce a collaborative audio-visual learning block to efficiently integrate multi-modal information while learning the visual and audio features. By iteratively employing this block, our single-stream network achieves a continuous fusion of multi-modal features across its layers. Thus, our network efficiently captures visual and audio features without the need for excessive block stacking, resulting in a lightweight network design. Furthermore, we propose a multi-modal classification module that can boost the dependence of the visual and audio classifiers on modality content. It also enhances the whole resistance of the video classifier against the mismatches between audio and visual modalities. We conduct experiments on the DF-TIMIT, FakeAVCeleb, and DFDC benchmark datasets. Compared to state-of-the-art audio-visual joint detection methods, our method is significantly lightweight with only 0.48M parameters, yet it achieves superiority in both uni-modal and multi-modal deepfakes, as well as in unseen types of deepfakes.
Semantic segmentation is one of the most fundamental tasks in image understanding with a long history of research, and subsequently a myriad of different approaches. Traditional methods strive to train models up from scratch, requiring vast amounts of computational resources and training data. In the advent of moving to open-vocabulary semantic segmentation, which asks models to classify beyond learned categories, large quantities of finely annotated data would be prohibitively expensive. Researchers have instead turned to training-free methods where they leverage existing models made for tasks where data is more easily acquired. Specifically, this survey will cover the history, nuance, idea development and the state-of-the-art in training-free open-vocabulary semantic segmentation that leverages existing multi-modal classification models. We will first give a preliminary on the task definition followed by an overview of popular model archetypes and then spotlight over 30 approaches split into broader research branches: purely CLIP-based, those leveraging auxiliary visual foundation models and ones relying on generative methods. Subsequently, we will discuss the limitations and potential problems of current research, as well as provide some underexplored ideas for future study. We believe this survey will serve as a good onboarding read to new researchers and spark increased interest in the area.
Human activity recognition (HAR) is essential for effective Human-Robot Collaboration (HRC), enabling robots to interpret and respond to human actions. This study evaluates the ability of a vision-based tactile sensor to classify 15 activities, comparing its performance to an IMU-based data glove. Additionally, we propose a multi-modal framework combining tactile and motion data to leverage their complementary strengths. We examined three approaches: motion-based classification (MBC) using IMU data, tactile-based classification (TBC) with single or dual video streams, and multi-modal classification (MMC) integrating both. Offline validation on segmented datasets assessed each configuration's accuracy under controlled conditions, while online validation on continuous action sequences tested online performance. Results showed the multi-modal approach consistently outperformed single-modality methods, highlighting the potential of integrating tactile and motion sensing to enhance HAR systems for collaborative robotics.
Accurate assessment of forest biodiversity is crucial for ecosystem management and conservation. While traditional field surveys provide high-quality assessments, they are labor-intensive and spatially limited. This study investigates whether deep learning-based fusion of close-range sensing data from 2D orthophotos (12.5 cm resolution) and 3D airborne laser scanning (ALS) point clouds (8 points/m^2) can enhance biodiversity assessment. We introduce the BioVista dataset, comprising 44.378 paired samples of orthophotos and ALS point clouds from temperate forests in Denmark, designed to explore multi-modal fusion approaches for biodiversity potential classification. Using deep neural networks (ResNet for orthophotos and PointVector for ALS point clouds), we investigate each data modality's ability to assess forest biodiversity potential, achieving mean accuracies of 69.4% and 72.8%, respectively. We explore two fusion approaches: a confidence-based ensemble method and a feature-level concatenation strategy, with the latter achieving a mean accuracy of 75.5%. Our results demonstrate that spectral information from orthophotos and structural information from ALS point clouds effectively complement each other in forest biodiversity assessment.
Multimodal AI models are increasingly used in fields like healthcare, finance, and autonomous driving, where information is drawn from multiple sources or modalities such as images, texts, audios, videos. However, effectively managing uncertainty - arising from noise, insufficient evidence, or conflicts between modalities - is crucial for reliable decision-making. Current uncertainty-aware machine learning methods leveraging, for example, evidence averaging, or evidence accumulation underestimate uncertainties in high-conflict scenarios. Moreover, the state-of-the-art evidence averaging strategy is not order invariant and fails to scale to multiple modalities. To address these challenges, we propose a novel multimodal learning method with order-invariant evidence fusion and introduce a conflict-based discounting mechanism that reallocates uncertain mass when unreliable modalities are detected. We provide both theoretical analysis and experimental validation, demonstrating that unlike the previous work, the proposed approach effectively distinguishes between conflicting and non-conflicting samples based on the provided uncertainty estimates, and outperforms the previous models in uncertainty-based conflict detection.
The prevalence of multi-modal content on social media complicates automated moderation strategies. This calls for an enhancement in multi-modal classification and a deeper understanding of understated meanings in images and memes. Although previous efforts have aimed at improving model performance through fine-tuning, few have explored an end-to-end optimization pipeline that accounts for modalities, prompting, labeling, and fine-tuning. In this study, we propose an end-to-end conceptual framework for model optimization in complex tasks. Experiments support the efficacy of this traditional yet novel framework, achieving the highest accuracy and AUROC. Ablation experiments demonstrate that isolated optimizations are not ineffective on their own.
Contrastive learning has become one of the most impressive approaches for multi-modal representation learning. However, previous multi-modal works mainly focused on cross-modal understanding, ignoring in-modal contrastive learning, which limits the representation of each modality. In this paper, we propose a novel contrastive learning strategy, called $Turbo$, to promote multi-modal understanding by joint in-modal and cross-modal contrastive learning. Specifically, multi-modal data pairs are sent through the forward pass twice with different hidden dropout masks to get two different representations for each modality. With these representations, we obtain multiple in-modal and cross-modal contrastive objectives for training. Finally, we combine the self-supervised Turbo with the supervised multi-modal classification and demonstrate its effectiveness on two audio-text classification tasks, where the state-of-the-art performance is achieved on a speech emotion recognition benchmark dataset.
We introduce a new, challenging benchmark and a dataset, FungiTastic, based on fungal records continuously collected over a twenty-year span. The dataset is labeled and curated by experts and consists of about 350k multimodal observations of 5k fine-grained categories (species). The fungi observations include photographs and additional data, e.g., meteorological and climatic data, satellite images, and body part segmentation masks. FungiTastic is one of the few benchmarks that include a test set with DNA-sequenced ground truth of unprecedented label reliability. The benchmark is designed to support (i) standard closed-set classification, (ii) open-set classification, (iii) multi-modal classification, (iv) few-shot learning, (v) domain shift, and many more. We provide baseline methods tailored for many use-cases, a multitude of ready-to-use pre-trained models on HuggingFace and a framework for model training. A comprehensive documentation describing the dataset features and the baselines are available at https://bohemianvra.github.io/FungiTastic/ and https://www.kaggle.com/datasets/picekl/fungitastic.
Vision-language models have emerged as a powerful tool for previously challenging multi-modal classification problem in the medical domain. This development has led to the exploration of automated image description generation for multi-modal clinical scans, particularly for radiology report generation. Existing research has focused on clinical descriptions for specific modalities or body regions, leaving a gap for a model providing entire-body multi-modal descriptions. In this paper, we address this gap by automating the generation of standardized body station(s) and list of organ(s) across the whole body in multi-modal MR and CT radiological images. Leveraging the versatility of the Contrastive Language-Image Pre-training (CLIP), we refine and augment the existing approach through multiple experiments, including baseline model fine-tuning, adding station(s) as a superset for better correlation between organs, along with image and language augmentations. Our proposed approach demonstrates 47.6% performance improvement over baseline PubMedCLIP.
Most convolutional neural network (CNN) based methods for skin cancer classification obtain their results using only dermatological images. Although good classification results have been shown, more accurate results can be achieved by considering the patient's metadata, which is valuable clinical information for dermatologists. Current methods only use the simple joint fusion structure (FS) and fusion modules (FMs) for the multi-modal classification methods, there still is room to increase the accuracy by exploring more advanced FS and FM. Therefore, in this paper, we design a new fusion method that combines dermatological images (dermoscopy images or clinical images) and patient metadata for skin cancer classification from the perspectives of FS and FM. First, we propose a joint-individual fusion (JIF) structure that learns the shared features of multi-modality data and preserves specific features simultaneously. Second, we introduce a fusion attention (FA) module that enhances the most relevant image and metadata features based on both the self and mutual attention mechanism to support the decision-making pipeline. We compare the proposed JIF-MMFA method with other state-of-the-art fusion methods on three different public datasets. The results show that our JIF-MMFA method improves the classification results for all tested CNN backbones and performs better than the other fusion methods on the three public datasets, demonstrating our method's effectiveness and robustness
In a joint vision-language space, a text feature (e.g., from "a photo of a dog") could effectively represent its relevant image features (e.g., from dog photos). Also, a recent study has demonstrated the cross-modal transferability phenomenon of this joint space. From these observations, we propose PromptStyler which simulates various distribution shifts in the joint space by synthesizing diverse styles via prompts without using any images to deal with source-free domain generalization. The proposed method learns to generate a variety of style features (from "a S* style of a") via learnable style word vectors for pseudo-words S*. To ensure that learned styles do not distort content information, we force style-content features (from "a S* style of a [class]") to be located nearby their corresponding content features (from "[class]") in the joint vision-language space. After learning style word vectors, we train a linear classifier using synthesized style-content features. PromptStyler achieves the state of the art on PACS, VLCS, OfficeHome and DomainNet, even though it does not require any images for training.
In the fashion domain, there exists a variety of vision-and-language (V+L) tasks, including cross-modal retrieval, text-guided image retrieval, multi-modal classification, and image captioning. They differ drastically in each individual input/output format and dataset size. It has been common to design a task-specific model and fine-tune it independently from a pre-trained V+L model (e.g., CLIP). This results in parameter inefficiency and inability to exploit inter-task relatedness. To address such issues, we propose a novel FAshion-focused Multi-task Efficient learning method for Vision-and-Language tasks (FAME-ViL) in this work. Compared with existing approaches, FAME-ViL applies a single model for multiple heterogeneous fashion tasks, therefore being much more parameter-efficient. It is enabled by two novel components: (1) a task-versatile architecture with cross-attention adapters and task-specific adapters integrated into a unified V+L model, and (2) a stable and effective multi-task training strategy that supports learning from heterogeneous data and prevents negative transfer. Extensive experiments on four fashion tasks show that our FAME-ViL can save 61.5% of parameters over alternatives, while significantly outperforming the conventional independently trained single-task models. Code is available at https://github.com/BrandonHanx/FAME-ViL.
In this paper, we first extend the recent Masked Auto-Encoder (MAE) model from a single modality to audio-visual multi-modalities. Subsequently, we propose the Contrastive Audio-Visual Masked Auto-Encoder (CAV-MAE) by combining contrastive learning and masked data modeling, two major self-supervised learning frameworks, to learn a joint and coordinated audio-visual representation. Our experiments show that the contrastive audio-visual correspondence learning objective not only enables the model to perform audio-visual retrieval tasks, but also helps the model learn a better joint representation. As a result, our fully self-supervised pretrained CAV-MAE achieves a new SOTA accuracy of 65.9% on VGGSound, and is comparable with the previous best supervised pretrained model on AudioSet in the audio-visual event classification task. Code and pretrained models are at https://github.com/yuangongnd/cav-mae.
Conventional audio-visual models have independent audio and video branches. In this work, we unify the audio and visual branches by designing a Unified Audio-Visual Model (UAVM). The UAVM achieves a new state-of-the-art audio-visual event classification accuracy of 65.8% on VGGSound. More interestingly, we also find a few intriguing properties of UAVM that the modality-independent counterparts do not have.
The automatic early diagnosis of prodromal stages of Alzheimer's disease is of great relevance for patient treatment to improve quality of life. We address this problem as a multi-modal classification task. Multi-modal data provides richer and complementary information. However, existing techniques only consider either lower order relations between the data and single/multi-modal imaging data. In this work, we introduce a novel semi-supervised hypergraph learning framework for Alzheimer's disease diagnosis. Our framework allows for higher-order relations among multi-modal imaging and non-imaging data whilst requiring a tiny labelled set. Firstly, we introduce a dual embedding strategy for constructing a robust hypergraph that preserves the data semantics. We achieve this by enforcing perturbation invariance at the image and graph levels using a contrastive based mechanism. Secondly, we present a dynamically adjusted hypergraph diffusion model, via a semi-explicit flow, to improve the predictive uncertainty. We demonstrate, through our experiments, that our framework is able to outperform current techniques for Alzheimer's disease diagnosis.
Making each modality in multi-modal data contribute is of vital importance to learning a versatile multi-modal model. Existing methods, however, are often dominated by one or few of modalities during model training, resulting in sub-optimal performance. In this paper, we refer to this problem as modality bias and attempt to study it in the context of multi-modal classification systematically and comprehensively. After stepping into several empirical analysis, we recognize that one modality affects the model prediction more just because this modality has a spurious correlation with instance labels. In order to primarily facilitate the evaluation on the modality bias problem, we construct two datasets respectively for the colored digit recognition and video action recognition tasks in line with the Out-of-Distribution (OoD) protocol. Collaborating with the benchmarks in the visual question answering task, we empirically justify the performance degradation of the existing methods on these OoD datasets, which serves as evidence to justify the modality bias learning. In addition, to overcome this problem, we propose a plug-and-play loss function method, whereby the feature space for each label is adaptively learned according to the training set statistics. Thereafter, we apply this method on eight baselines in total to test its effectiveness. From the results on four datasets regarding the above three tasks, our method yields remarkable performance improvements compared with the baselines, demonstrating its superiority on reducing the modality bias problem.
Sarcasm detection and humor classification are inherently subtle problems, primarily due to their dependence on the contextual and non-verbal information. Furthermore, existing studies in these two topics are usually constrained in non-English languages such as Hindi, due to the unavailability of qualitative annotated datasets. In this work, we make two major contributions considering the above limitations: (1) we develop a Hindi-English code-mixed dataset, MaSaC, for the multi-modal sarcasm detection and humor classification in conversational dialog, which to our knowledge is the first dataset of its kind; (2) we propose MSH-COMICS, a novel attention-rich neural architecture for the utterance classification. We learn efficient utterance representation utilizing a hierarchical attention mechanism that attends to a small portion of the input sentence at a time. Further, we incorporate dialog-level contextual attention mechanism to leverage the dialog history for the multi-modal classification. We perform extensive experiments for both the tasks by varying multi-modal inputs and various submodules of MSH-COMICS. We also conduct comparative analysis against existing approaches. We observe that MSH-COMICS attains superior performance over the existing models by > 1 F1-score point for the sarcasm detection and 10 F1-score points in humor classification. We diagnose our model and perform thorough analysis of the results to understand the superiority and pitfalls.
Recent advances in using retrieval components over external knowledge sources have shown impressive results for a variety of downstream tasks in natural language processing. Here, we explore the use of unstructured external knowledge sources of images and their corresponding captions for improving visual question answering (VQA). First, we train a novel alignment model for embedding images and captions in the same space, which achieves substantial improvement in performance on image-caption retrieval w.r.t. similar methods. Second, we show that retrieval-augmented multi-modal transformers using the trained alignment model improve results on VQA over strong baselines. We further conduct extensive experiments to establish the promise of this approach, and examine novel applications for inference time such as hot-swapping indices.
While many approaches exist in the literature to learn low-dimensional representations for data collections in multiple modalities, the generalizability of multi-modal nonlinear embeddings to previously unseen data is a rather overlooked subject. In this work, we first present a theoretical analysis of learning multi-modal nonlinear embeddings in a supervised setting. Our performance bounds indicate that for successful generalization in multi-modal classification and retrieval problems, the regularity of the interpolation functions extending the embedding to the whole data space is as important as the between-class separation and cross-modal alignment criteria. We then propose a multi-modal nonlinear representation learning algorithm that is motivated by these theoretical findings, where the embeddings of the training samples are optimized jointly with the Lipschitz regularity of the interpolators. Experimental comparison to recent multi-modal and single-modal learning algorithms suggests that the proposed method yields promising performance in multi-modal image classification and cross-modal image-text retrieval applications.
Compared to natural images, understanding scientific figures is particularly hard for machines. However, there is a valuable source of information in scientific literature that until now has remained untapped: the correspondence between a figure and its caption. In this paper we investigate what can be learnt by looking at a large number of figures and reading their captions, and introduce a figure-caption correspondence learning task that makes use of our observations. Training visual and language networks without supervision other than pairs of unconstrained figures and captions is shown to successfully solve this task. We also show that transferring lexical and semantic knowledge from a knowledge graph significantly enriches the resulting features. Finally, we demonstrate the positive impact of such features in other tasks involving scientific text and figures, like multi-modal classification and machine comprehension for question answering, outperforming supervised baselines and ad-hoc approaches.