We present a framework for learning multimodal representations from unlabeled data using convolution-free Transformer architectures. Specifically, our Video-Audio-Text Transformer (VATT) takes raw signals as inputs and extracts multimodal representations that are rich enough to benefit a variety of downstream tasks. We train VATT end-to-end from scratch using multimodal contrastive losses and evaluate its performance by the downstream tasks of video action recognition, audio event classification, image classification, and text-to-video retrieval. Furthermore, we study a modality-agnostic single-backbone Transformer by sharing weights among the three modalities. We show that the convolution-free VATT outperforms state-of-the-art ConvNet-based architectures in the downstream tasks. Especially, VATT's vision Transformer achieves the top-1 accuracy of 82.1% on Kinetics-400, 83.6% on Kinetics-600,and 41.1% on Moments in Time, new records while avoiding supervised pre-training. Transferring to image classification leads to 78.7% top-1 accuracy on ImageNet compared to 64.7% by training the same Transformer from scratch, showing the generalizability of our model despite the domain gap between videos and images. VATT's audio Transformer also sets a new record on waveform-based audio event recognition by achieving the mAP of 39.4% on AudioSet without any supervised pre-training.

Few-shot object detection (FSOD) aims to detect objects using only few examples. It's critically needed for many practical applications but so far remains challenging. We propose a meta-learning based few-shot object detection method by transferring meta-knowledge learned from data-abundant base classes to data-scarce novel classes. Our method incorporates a coarse-to-fine approach into the proposal based object detection framework and integrates prototype based classifiers into both the proposal generation and classification stages. To improve proposal generation for few-shot novel classes, we propose to learn a lightweight matching network to measure the similarity between each spatial position in the query image feature map and spatially-pooled class features, instead of the traditional object/nonobject classifier, thus generating category-specific proposals and improving proposal recall for novel classes. To address the spatial misalignment between generated proposals and few-shot class examples, we propose a novel attentive feature alignment method, thus improving the performance of few-shot object detection. Meanwhile we jointly learn a Faster R-CNN detection head for base classes. Extensive experiments conducted on multiple FSOD benchmarks show our proposed approach achieves state of the art results under (incremental) few-shot learning settings.

Despite the remarkable accuracy of deep neural networks in object detection, they are costly to train and scale due to supervision requirements. Particularly, learning more object categories typically requires proportionally more bounding box annotations. Weakly supervised and zero-shot learning techniques have been explored to scale object detectors to more categories with less supervision, but they have not been as successful and widely adopted as supervised models. In this paper, we put forth a novel formulation of the object detection problem, namely open-vocabulary object detection, which is more general, more practical, and more effective than weakly supervised and zero-shot approaches. We propose a new method to train object detectors using bounding box annotations for a limited set of object categories, as well as image-caption pairs that cover a larger variety of objects at a significantly lower cost. We show that the proposed method can detect and localize objects for which no bounding box annotation is provided during training, at a significantly higher accuracy than zero-shot approaches. Meanwhile, objects with bounding box annotation can be detected almost as accurately as supervised methods, which is significantly better than weakly supervised baselines. Accordingly, we establish a new state of the art for scalable object detection.

Neuro-symbolic representations have proved effective in learning structure information in vision and language. In this paper, we propose a new model architecture for learning multi-modal neuro-symbolic representations for video captioning. Our approach uses a dictionary learning-based method of learning relations between videos and their paired text descriptions. We refer to these relations as relative roles and leverage them to make each token role-aware using attention. This results in a more structured and interpretable architecture that incorporates modality-specific inductive biases for the captioning task. Intuitively, the model is able to learn spatial, temporal, and cross-modal relations in a given pair of video and text. The disentanglement achieved by our proposal gives the model more capacity to capture multi-modal structures which result in captions with higher quality for videos. Our experiments on two established video captioning datasets verifies the effectiveness of the proposed approach based on automatic metrics. We further conduct a human evaluation to measure the grounding and relevance of the generated captions and observe consistent improvement for the proposed model. The codes and trained models can be found at https://github.com/hassanhub/R3Transformer

The prevailing framework for solving referring expression grounding is based on a two-stage process: 1) detecting proposals with an object detector and 2) grounding the referent to one of the proposals. Existing two-stage solutions mostly focus on the grounding step, which aims to align the expressions with the proposals. In this paper, we argue that these methods overlook an obvious mismatch between the roles of proposals in the two stages: they generate proposals solely based on the detection confidence (i.e., expression-agnostic), hoping that the proposals contain all right instances in the expression (i.e., expression-aware). Due to this mismatch, current two-stage methods suffer from a severe performance drop between detected and ground-truth proposals. To this end, we propose Ref-NMS, which is the first method to yield expression-aware proposals at the first stage. Ref-NMS regards all nouns in the expression as critical objects, and introduces a lightweight module to predict a score for aligning each box with a critical object. These scores can guide the NMS operation to filter out the boxes irrelevant to the expression, increasing the recall of critical objects, resulting in a significantly improved grounding performance. Since Ref-NMS is agnostic to the grounding step, it can be easily integrated into any state-of-the-art two-stage method. Extensive ablation studies on several backbones, benchmarks, and tasks consistently demonstrate the superiority of Ref-NMS.

Children acquire language subconsciously by observing the surrounding world and listening to descriptions. They can discover the meaning of words even without explicit language knowledge, and generalize to novel compositions effortlessly. In this paper, we bring this ability to AI, by studying the task of Visually grounded Language Acquisition (VLA). We propose a multimodal transformer model augmented with a novel mechanism for analogical reasoning, which approximates novel compositions by learning semantic mapping and reasoning operations from previously seen compositions. Our proposed method, Analogical Reasoning Transformer Networks (ARTNet), is trained on raw multimedia data (video frames and transcripts), and after observing a set of compositions such as "washing apple" or "cutting carrot", it can generalize and recognize new compositions in new video frames, such as "washing carrot" or "cutting apple". To this end, ARTNet refers to relevant instances in the training data and uses their visual features and captions to establish analogies with the query image. Then it chooses the suitable verb and noun to create a new composition that describes the new image best. Extensive experiments on an instructional video dataset demonstrate that the proposed method achieves significantly better generalization capability and recognition accuracy compared to state-of-the-art transformer models.

To combat COVID-19, both clinicians and scientists need to digest the vast amount of relevant biomedical knowledge in literature to understand the disease mechanism and the related biological functions. We have developed a novel and comprehensive knowledge discovery framework, \textbf{COVID-KG} to extract fine-grained multimedia knowledge elements (entities, relations and events) from scientific literature. We then exploit the constructed multimedia knowledge graphs (KGs) for question answering and report generation, using drug repurposing as a case study. Our framework also provides detailed contextual sentences, subfigures and knowledge subgraphs as evidence. All of the data, KGs, reports, resources and shared services are publicly available.

Many real-world applications involve multivariate, geo-tagged time series data: at each location, multiple sensors record corresponding measurements. For example, air quality monitoring system records PM2.5, CO, etc. The resulting time-series data often has missing values due to device outages or communication errors. In order to impute the missing values, state-of-the-art methods are built on Recurrent Neural Networks (RNN), which process each time stamp sequentially, prohibiting the direct modeling of the relationship between distant time stamps. Recently, the self-attention mechanism has been proposed for sequence modeling tasks such as machine translation, significantly outperforming RNN because the relationship between each two time stamps can be modeled explicitly. In this paper, we are the first to adapt the self-attention mechanism for multivariate, geo-tagged time series data. In order to jointly capture the self-attention across multiple dimensions, including time, location and the sensor measurements, while maintain low computational complexity, we propose a novel approach called Cross-Dimensional Self-Attention (CDSA) to process each dimension sequentially, yet in an order-independent manner. Our extensive experiments on four real-world datasets, including three standard benchmarks and our newly collected NYC-traffic dataset, demonstrate that our approach outperforms the state-of-the-art imputation and forecasting methods. A detailed systematic analysis confirms the effectiveness of our design choices.

Deep neural networks suffer from over-fitting and catastrophic forgetting when trained with small data. One natural remedy for this problem is data augmentation, which has been recently shown to be effective. However, previous works either assume that intra-class variances can always be generalized to new classes, or employ naive generation methods to hallucinate finite examples without modeling their latent distributions. In this work, we propose Covariance-Preserving Adversarial Augmentation Networks to overcome existing limits of low-shot learning. Specifically, a novel Generative Adversarial Network is designed to model the latent distribution of each novel class given its related base counterparts. Since direct estimation on novel classes can be inductively biased, we explicitly preserve covariance information as the ``variability'' of base examples during the generation process. Empirical results show that our model can generate realistic yet diverse examples, leading to substantial improvements on the ImageNet benchmark over the state of the art.

Deep neural networks suffer from over-fitting and catastrophic forgetting when trained with small data. One natural remedy for this problem is data augmentation, which has been recently shown to be effective. However, previous works either assume that intra-class variances can always be generalized to new classes, or employ naive generation methods to hallucinate finite examples without modeling their latent distributions. In this work, we propose Covariance-Preserving Adversarial Augmentation Networks to overcome existing limits of low-shot learning. Specifically, a novel Generative Adversarial Network is designed to model the latent distribution of each novel class given its related base counterparts. Since direct estimation on novel classes can be inductively biased, we explicitly preserve covariance information as the ``variability'' of base examples during the generation process. Empirical results show that our model can generate realistic yet diverse examples, leading to substantial improvements on the ImageNet benchmark over the state of the art.