Neural Information Processing Systems http://nips.cc/

We extend the popular BERT architecture to a multi-modal two-stream model, processing both visual and textual inputs in separate streams that interact through co-attentional transformer layers. We pretrain our model through two proxy tasks on the large, automatically collected Conceptual Captions dataset and then transfer it to multiple established vision-and-language tasks -- visual question answering, visual commonsense reasoning, referring expressions, and caption-based image retrieval -- by making only minor additions to the base architecture. We observe significant improvements across tasks compared to existing task-specific models -- achieving state-of-the-art on all four tasks. Our work represents a shift away from learning groundings between vision and language only as part of task training and towards treating visual grounding as a pretrainable and transferable capability.

Current state-of-the-art vision-and-language models are evaluated on tasks either individually or in a multi-task setting, overlooking the challenges of continually learning (CL) tasks as they arrive. Existing CL benchmarks have facilitated research on task adaptation and mitigating catastrophic forgetting, but are limited to vision-only and language-only tasks. We present CLiMB, a benchmark to study the challenge of learning multimodal tasks in a CL setting, and to systematically evaluate how upstream continual learning can rapidly generalize to new multimodal and unimodal tasks. CLiMB includes implementations of several CL algorithms and a modified Vision-Language Transformer (ViLT) model that can be deployed on both multimodal and unimodal tasks. We find that common CL methods can help mitigate forgetting during multimodal task learning, but do not enable cross-task knowledge transfer. We envision that CLiMB will facilitate research on a new class of CL algorithms for this challenging multimodal setting.

Neural Information Processing Systems http://nips.cc/

I think that this paper is a solid extension of masked language model pre-training to image-and-text (e.g., captioning) tasks. It defines two novel but intuitive pre-training tasks for this scenario: (i) predicting the semantic class of masked image regions given the surrounding image regions (from the same image) and the corresponding text, (ii) predicting whether image and text pairs are aligned. They demonstrate significant improvements over both the previous SOTA and the strong baseline of simply using a pre-trained text-only BERT model. They also show that having two encoders (with different parameters), one for images and one for text, is superior to a joint encoder. I would have liked to have seen more ablation of the pre-training tasks, since I think that this is more interesting than the model depth ablation that the authors performed.

Current state-of-the-art vision-and-language models are evaluated on tasks either individually or in a multi-task setting, overlooking the challenges of continually learning (CL) tasks as they arrive. Existing CL benchmarks have facilitated research on task adaptation and mitigating "catastrophic forgetting", but are limited to vision-only and language-only tasks. We present CLiMB, a benchmark to study the challenge of learning multimodal tasks in a CL setting, and to systematically evaluate how upstream continual learning can rapidly generalize to new multimodal and unimodal tasks. CLiMB includes implementations of several CL algorithms and a modified Vision-Language Transformer (ViLT) model that can be deployed on both multimodal and unimodal tasks. We find that common CL methods can help mitigate forgetting during multimodal task learning, but do not enable cross-task knowledge transfer.

We extend the popular BERT architecture to a multi-modal two-stream model, processing both visual and textual inputs in separate streams that interact through co-attentional transformer layers. We pretrain our model through two proxy tasks on the large, automatically collected Conceptual Captions dataset and then transfer it to multiple established vision-and-language tasks -- visual question answering, visual commonsense reasoning, referring expressions, and caption-based image retrieval -- by making only minor additions to the base architecture. We observe significant improvements across tasks compared to existing task-specific models -- achieving state-of-the-art on all four tasks. Our work represents a shift away from learning groundings between vision and language only as part of task training and towards treating visual grounding as a pretrainable and transferable capability.

As able-bodied people, we often take our vision for granted. For people who are visually impaired, however, their disability can have a significant impact on their daily lives. We are developing proprietary headgear that will help visually impaired people navigate their surroundings, identify objects and people, read text, and avoid obstacles. The headgear will use a combination of computer vision, distance estimation with ultrasonic sensors, voice recognition, and voice assistants to provide users with real-time information about their environment. Users will be able to interact with the headgear through voice commands, such as ''What is that?'' to identify an object or ''Navigate to the front door'' to find their way around. The headgear will then provide the user with a verbal description of the object or spoken navigation instructions. We believe that this headgear has the potential to make a significant difference in the lives of visually impaired people, allowing them to live more independently and participate more fully in society.

This paper surveys vision-language pre-training (VLP) methods for multimodal intelligence that have been developed in the last few years. We group these approaches into three categories: ($i$) VLP for image-text tasks, such as image captioning, image-text retrieval, visual question answering, and visual grounding; ($ii$) VLP for core computer vision tasks, such as (open-set) image classification, object detection, and segmentation; and ($iii$) VLP for video-text tasks, such as video captioning, video-text retrieval, and video question answering. For each category, we present a comprehensive review of state-of-the-art methods, and discuss the progress that has been made and challenges still being faced, using specific systems and models as case studies. In addition, for each category, we discuss advanced topics being actively explored in the research community, such as big foundation models, unified modeling, in-context few-shot learning, knowledge, robustness, and computer vision in the wild, to name a few.

Is more data always better to train vision-and-language models? We study knowledge transferability in multi-modal tasks. The current tendency in machine learning is to assume that by joining multiple datasets from different tasks their overall performance will improve. However, we show that not all the knowledge transfers well or has a positive impact on related tasks, even when they share a common goal. We conduct an exhaustive analysis based on hundreds of cross-experiments on 12 vision-and-language tasks categorized in 4 groups. Whereas tasks in the same group are prone to improve each other, results show that this is not always the case. Other factors such as dataset size or pre-training stage have also a great impact on how well the knowledge is transferred.