MAMO: Fine-Grained Vision-Language Representations Learning with Masked Multimodal Modeling

Deep learning methods have revolutionized speech recognition, image recognition, and natural language processing since 2010. Each of these tasks involves a single modality in their input signals. However, many applications in the artificial intelligence field involve multiple modalities. Therefore, it is of broad interest to study the more difficult and complex problem of modeling and learning across multiple modalities. In this paper, we provide a technical review of available models and learning methods for multimodal intelligence. The main focus of this review is the combination of vision and natural language modalities, which has become an important topic in both the computer vision and natural language processing research communities. This review provides a comprehensive analysis of recent works on multimodal deep learning from three perspectives: learning multimodal representations, fusing multimodal signals at various levels, and multimodal applications. Regarding multimodal representation learning, we review the key concepts of embedding, which unify multimodal signals into a single vector space and thereby enable cross-modality signal processing. We also review the properties of many types of embeddings that are constructed and learned for general downstream tasks. Regarding multimodal fusion, this review focuses on special architectures for the integration of representations of unimodal signals for a particular task. Regarding applications, selected areas of a broad interest in the current literature are covered, including image-to-text caption generation, text-to-image generation, and visual question answering. We believe that this review will facilitate future studies in the emerging field of multimodal intelligence for related communities.

Multimodal representations play an important role in multimodal learning tasks, including cross-modal retrieval and intra-modal clustering. However, existing multimodal representation learning approaches focus on building one common space by aligning different modalities and ignore the complementary information across the modalities, such as the intra-modal local structures. In other words, they only focus on the object-level alignment and ignore structure-level alignment. To tackle the problem, we propose a novel symmetric multimodal representation learning framework by transferring local structures across different modalities, namely MTLS. A customized soft metric learning strategy and an iterative parameter learning process are designed to symmetrically transfer local structures and enhance the cluster structures in intra-modal representations. The bidirectional retrieval loss based on multi-layer neural networks is utilized to align two modalities. MTLS is instantiated with image and text data and shows its superior performance on image-text retrieval and image clustering. MTLS outperforms the state-of-the-art multimodal learning methods by up to 32% in terms of R@1 on text-image retrieval and 16.4% in terms of AMI onclustering.

Data augmentation is a necessity to enhance data efficiency in deep learning. For vision-language pre-training, data is only augmented either for images or for text in previous works. In this paper, we present MixGen: a joint data augmentation for vision-language representation learning to further improve data efficiency. It generates new image-text pairs with semantic relationships preserved by interpolating images and concatenating text. It's simple, and can be plug-and-played into existing pipelines. We evaluate MixGen on four architectures, including CLIP, ViLT, ALBEF and TCL, across five downstream vision-language tasks to show its versatility and effectiveness. For example, adding MixGen in ALBEF pre-training leads to absolute performance improvements on downstream tasks: image-text retrieval (+6.2% on COCO fine-tuned and +5.3% on Flicker30K zero-shot), visual grounding (+0.9% on Re-fCOCO+), visual reasoning (+0.9% on NLVR <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ), visual question answering (+0.3% on VQA2.0), and visual entail-ment (+0.4% on SNLI-VE).

Multi-modal knowledge graphs representation learning via multi-headed self-attention

Vision-language representation learning largely benefits from image-text alignment through contrastive losses (e.g., InfoNCE loss). The success of this alignment strategy is attributed to its capability in maximizing the mutual information (MI) between an image and its matched text. However, simply performing cross-modal alignment (CMA) ignores data potential within each modality, which may result in degraded representations. For instance, although CMA-based models are able to map image-text pairs close together in the embedding space, they fail to ensure that similar inputs from the same modality stay close by. This problem can get even worse when the pre-training data is noisy. In this paper, we propose triple contrastive learning (TCL) for vision-language pre-training by leveraging both cross-modal and intra-modal self-supervision. Besides CMA, TCL introduces an intra-modal contrastive objective to provide complementary benefits in representation learning. To take advantage of localized and structural information from image and text input, TCL further maximizes the average MI between local regions of image/text and their global summary. To the best of our knowledge, ours is the first work that takes into account local structure information for multi-modality representation learning. Experimental evaluations show that our approach is competitive and achieves the new state of the art on various common downstream vision-language tasks such as image-text retrieval and visual question answering.

The combination of visual and textual representations has produced excellent results in tasks such as image captioning and visual question answering, but the inference capabilities of multimodal representations are largely untested. In the case of textual representations, inference tasks such as Textual Entailment and Semantic Textual Similarity have been often used to benchmark the quality of textual representations. The long term goal of our research is to devise multimodal representation techniques that improve current inference capabilities. We thus present a novel task, Visual Semantic Textual Similarity (vSTS), where such inference ability can be tested directly. Given two items comprised each by an image and its accompanying caption, vSTS systems need to assess the degree to which the captions in context are semantically equivalent to each other. Our experiments using simple multimodal representations show that the addition of image representations produces better inference, compared to text-only representations. The improvement is observed both when directly computing the similarity between the representations of the two items, and when learning a siamese network based on vSTS training data. Our work shows, for the first time, the successful contribution of visual information to textual inference, with ample room for benchmarking more complex multimodal representation options.

Shouldn't language and vision features be treated equally in vision-language (VL) tasks? Many VL approaches treat the language component as an afterthought, using simple language models that are either built upon fixed word embeddings trained on text-only data or are learned from scratch. We conclude that language features deserve more attention, which has been informed by experiments which compare different word embeddings, language models, and embedding augmentation steps on five common VL tasks: image-sentence retrieval, image captioning, visual question answering, phrase grounding, and text-to-clip retrieval. Our experiments provide some striking results; an average embedding language model outperforms a LSTM on retrieval-style tasks; state-of-the-art representations such as BERT perform relatively poorly on vision-language tasks. From this comprehensive set of experiments we can propose a set of best practices for incorporating the language component of vision-language tasks. To further elevate language features, we also show that knowledge in vision-language problems can be transferred across tasks to gain performance with multi-task training. This multi-task training is applied to a new Graph Oriented Vision-Language Embedding (GrOVLE), which we adapt from Word2Vec using WordNet and an original visual-language graph built from Visual Genome, providing a ready-to-use vision-language embedding: http://ai.bu.edu/grovle.

Modular Arabic VQA System Using Pre-Trained Models

Multimodal feature fusion by relational reasoning and attention for visual question answering

We study the problem of visual question answering (VQA) in images by exploiting supervised domain adaptation, where there is a large amount of labeled data in the source domain but only limited labeled data in the target domain, with the goal to train a good target model. A straightforward solution is to fine-tune a pre-trained source model by using those limited labeled target data, but it usually cannot work well due to the considerable difference between the data distributions of the source and target domains. Moreover, the availability of multiple modalities (i.e., images, questions and answers) in VQA poses further challenges in modeling the transferability between various modalities. In this paper, we address the above issues by proposing a novel supervised multi-modal domain adaptation method for VQA to learn joint feature embeddings across different domains and modalities. Specifically, we align the data distributions of the source and target domains by considering those modalities both jointly and separately. Extensive experiments on the benchmark VQA 2.0 and VizWiz datasets demonstrate that our proposed method outperforms the existing state-of-the-art baselines for open-ended VQA in this challenging domain adaptation setting.

In this paper, we propose a method to identify identical commodities. In e-commerce scenarios, commodities are usually described by both images and text. By definition, identical commodities are those that have identical key attributes and are cognitively identical to consumers. There are two main challenges: 1) The extraction and fusion of multi-modal representation. 2) The ability to verify identical commodities by comparing the similarity between representations and a threshold. To address the above problems, we propose an end-to-end multi-modal representation learning method with self-adaptive threshold. We use a dual-stream network to extract multi-modal commodity embeddings and threshold embeddings separately and then concatenate them to obtain commodity representation. Our method is able to adaptively adjust the threshold according to different commodities while maintaining the indexability of the commodity representation space. We experimentally validate the advantages of self-adaptive threshold and the effectiveness of multimodal representation fusion. Besides, our method achieves third place with an F1 score of 0.8936 on the second task of the CCKS-2022 Knowledge Graph Evaluation for Digital Commerce Competition. Code and pretrained models are available at https://github.com/hanchenchen/CCKS2022-track2-solution.KeywordsMulti-modal representationSelf-adaptive thresholdCCKS-2022 competition

Toxicity is paramount for comprehending compound properties, particularly in the early stages of drug design. Due to the diversity and complexity of toxic effects, it became a challenge to compute compound toxicity tasks. To address this issue, we propose a multimodal representation learning model, termed multimodal graph isomorphism network (MMGIN), to address this challenge for compound toxicity multitask learning. Based on fingerprints and molecular graphs of compounds, our MMGIN model incorporates a multimodal representation learning model to acquire a comprehensive compound representation. This model adopts a two-channel structure to independently learn fingerprint representation and molecular graph representation. Subsequently, two feedforward neural networks utilize the learned multimodal compound representation to perform multitask learning, encompassing compound toxicity classification and multiple compound category classification simultaneously. To test the effectiveness of our model, we constructed a novel data set, termed the compound toxicity multitask learning (CTMTL) data set, derived from the TOXRIC data set. We compare our MMGIN model with other representative machine learning and deep learning models on the CTMTL and Tox21 data sets. The experimental results demonstrate significant advancements achieved by our MMGIN model. Furthermore, the ablation study underscores the effectiveness of the introduced fingerprints, molecular graphs, the multimodal representation learning model, and the multitask learning model, showcasing the model's superior predictive capability and robustness.

Visual Question Answering (VQA) is a method of answering questions in natural language based on the content of images and has been widely concerned by researchers. The existing research on the visual question answering model mainly focuses on the point of view of attention mechanism and multi-modal fusion. It only pays attention to the visual semantic features of the image in the process of image modeling, ignoring the importance of modeling the spatial relationship of visual objects. We are aiming at the existing problems of the existing VQA model research. An effective spatial relationship reasoning network model is proposed, which can combine visual object semantic reasoning and spatial relationship reasoning at the same time to realize fine-grained multi-modal reasoning and fusion. A sparse attention encoder is designed to capture contextual information effectively in the semantic reasoning module. In the spatial relationship reasoning module, the graph neural network attention mechanism is used to model the spatial relationship of visual objects, which can correctly answer complex spatial relationship reasoning questions. Finally, a practical compact self-attention (CSA) mechanism is designed to reduce the redundancy of self-attention in linear transformation and the number of model parameters and effectively improve the model's overall performance. Quantitative and qualitative experiments are conducted on the benchmark datasets of VQA 2.0 and GQA. The experimental results demonstrate that the proposed method performs favorably against the state-of-the-art approaches. Our best single model has an overall accuracy of 71.18% on the VQA 2.0 dataset and 57.59% on the GQA dataset.

Recent advances in vision-language models have resulted in improved accuracy in visual question answering (VQA) tasks. However, their robustness remains limited when faced with out-of-distribution data containing unanswerable questions. In this study, we first construct a simple randomised VQA dataset, incorporating unanswerable questions from the VQA v2 dataset, to evaluate the robustness of a state-of-the-art VQA model. Our findings reveal that the model struggles to predict the “unknown” answer or provides inaccurate responses with high confidence scores for irrelevant questions. To address this issue without retraining the large backbone models, we propose Cross Modal Augmentation (CMA), a model-agnostic, test-time-only, multi-modal semantic augmentation technique. CMA generates multiple semantically-consistent but heterogeneous instances from the visual and textual inputs, which are then fed to the model, and the predictions are combined to achieve a more robust output. We demonstrate that implementing CMA enables the VQA model to provide more reliable answers in scenarios involving unanswerable questions, and show that the approach is generalisable across different categories of pre-trained vision language models.

Schizophrenia is a severe psychiatric disorder associated with a wide range of cognitive and neurophysiological dysfunctions and long-term social difficulties. In this paper, we test the hypothesis that integration of multiple simultaneous acquisitions of neuroimaging, behavioral, and clinical information will be better for prediction of early psychosis than unimodal recordings. We propose a novel framework to investigate the neural underpinnings of the early psychosis symptoms (that can develop into Schizophrenia with age) using multimodal acquisitions of neural and behavioral recordings including functional near-infrared spectroscopy (fNIRS) and electroencephalography (EEG), and facial features. Our data acquisition paradigm is based on live face-to-face interaction in order to study the neural correlates of social cognition in first-episode psychosis (FEP). We propose a novel deep representation learning framework, Neural-PRISM, for learning joint multimodal compressed representations combining neural as well as behavioral recordings. These learned representations are subsequently used to describe, classify, and predict the severity of early psychosis in patients, as measured by the Positive and Negative Syndrome Scale (PANSS) and Global Assessment of Functioning (GAF) scores. We found that incorporating joint multimodal representations from fNIRS and EEG along with behavioral recordings enhances classification between typical controls and FEP individuals. Additionally, our results suggest that geometric and topological features such as curvatures and path signatures of the embedded trajectories of brain activity enable detection of discriminatory neural characteristics in early psychosis.