This work concerns video-language pre-training and representation learning. In this now ubiquitous training scheme, a model first performs pre-training on paired videos and text (e.g., video clips and accompanied subtitles) from a large uncurated source corpus, before transferring to specific downstream tasks. This two-stage training process inevitably raises questions about the generalization ability of the pre-trained model, which is particularly pronounced when a salient domain gap exists between source and target data (e.g., instructional cooking videos vs. movies). In this paper, we first bring to light the sensitivity of pre-training objectives (contrastive vs. reconstructive) to domain discrepancy. Then, we propose a simple yet effective framework, CUPID, to bridge this domain gap by filtering and adapting source data to the target data, followed by domain-focused pre-training. Comprehensive experiments demonstrate that pre-training on a considerably small subset of domain-focused data can effectively close the source-target domain gap and achieve significant performance gain, compared to random sampling or even exploiting the full pre-training dataset. CUPID yields new state-of-the-art performance across multiple video-language and video tasks, including text-to-video retrieval [72, 37], video question answering [36], and video captioning [72], with consistent performance lift over different pre-training methods.
Vision-and-language pre-training has achieved impressive success in learning multimodal representations between vision and language. To generalize this success to non-English languages, we introduce UC2, the first machine translation-augmented framework for cross-lingual cross-modal representation learning. To tackle the scarcity problem of multilingual captions for image datasets, we first augment existing English-only datasets with other languages via machine translation (MT). Then we extend the standard Masked Language Modeling and Image-Text Matching training objectives to multilingual setting, where alignment between different languages is captured through shared visual context (i.e, using image as pivot). To facilitate the learning of a joint embedding space of images and all languages of interest, we further propose two novel pre-training tasks, namely Masked Region-to-Token Modeling (MRTM) and Visual Translation Language Modeling (VTLM), leveraging MT-enhanced translated data. Evaluation on multilingual image-text retrieval and multilingual visual question answering benchmarks demonstrates that our proposed framework achieves new state-of-the-art on diverse non-English benchmarks while maintaining comparable performance to monolingual pre-trained models on English tasks.
The canonical approach to video-and-language learning (e.g., video question answering) dictates a neural model to learn from offline-extracted dense video features from vision models and text features from language models. These feature extractors are trained independently and usually on tasks different from the target domains, rendering these fixed features sub-optimal for downstream tasks. Moreover, due to the high computational overload of dense video features, it is often difficult (or infeasible) to plug feature extractors directly into existing approaches for easy finetuning. To provide a remedy to this dilemma, we propose a generic framework ClipBERT that enables affordable end-to-end learning for video-and-language tasks, by employing sparse sampling, where only a single or a few sparsely sampled short clips from a video are used at each training step. Experiments on text-to-video retrieval and video question answering on six datasets demonstrate that ClipBERT outperforms (or is on par with) existing methods that exploit full-length videos, suggesting that end-to-end learning with just a few sparsely sampled clips is often more accurate than using densely extracted offline features from full-length videos, proving the proverbial less-is-more principle. Videos in the datasets are from considerably different domains and lengths, ranging from 3-second generic domain GIF videos to 180-second YouTube human activity videos, showing the generalization ability of our approach. Comprehensive ablation studies and thorough analyses are provided to dissect what factors lead to this success. Our code is publicly available at https://github.com/jayleicn/ClipBERT
Articulated hand pose tracking is an underexplored problem that carries the potential for use in an extensive number of applications, especially in the medical domain. With a robust and accurate tracking system on in-vivo surgical videos, the motion dynamics and movement patterns of the hands can be captured and analyzed for rich tasks including skills assessment, training surgical residents, and temporal action recognition. In this work, we propose a novel hand pose estimation model, Res152- CondPose, which improves tracking accuracy by incorporating a hand pose prior into its pose prediction. We show improvements over state-of-the-art methods which provide frame-wise independent predictions, by following a temporally guided approach that effectively leverages past predictions. Additionally, we collect the first dataset, Surgical Hands, that provides multi-instance articulated hand pose annotations for in-vivo videos. Our dataset contains 76 video clips from 28 publicly available surgical videos and over 8.1k annotated hand pose instances. We provide bounding boxes, articulated hand pose annotations, and tracking IDs to enable multi-instance area-based and articulated tracking. When evaluated on Surgical Hands, we show our method outperforms the state-of-the-art method using mean Average Precision (mAP), to measure pose estimation accuracy, and Multiple Object Tracking Accuracy (MOTA), to assess pose tracking performance.
Transformer has become ubiquitous in the deep learning field. One of the key ingredients that destined its success is the self-attention mechanism, which allows fully-connected contextual encoding over input tokens. However, despite its effectiveness in modeling short sequences, self-attention suffers when handling inputs with extreme long-range dependencies, as its complexity grows quadratically with respect to the sequence length. Therefore, long sequences are often encoded by Transformer in chunks using a sliding window. In this paper, we propose Cluster-Former, a novel clustering-based sparse Transformer to perform attention across chunked sequences. Our proposed method allows information integration beyond local windows, which is especially beneficial for question answering (QA) and language modeling tasks that rely on long-range dependencies. Experiments show that Cluster-Former achieves state-of-the-art performance on several major QA benchmarks.
This paper presents a unified Vision-Language Pre-training (VLP) model. The model is unified in that (1) it can be fine-tuned for either vision-language generation (e.g., image captioning) or understanding (e.g., visual question answering) tasks, and (2) it uses a shared multi-layer transformer network for both encoding and decoding, which differs from many existing methods where the encoder and decoder are implemented using separate models. The unified VLP model is pre-trained on a large amount of image-text pairs using the unsupervised learning objectives of two tasks: bidirectional and sequence-to-sequence (seq2seq) masked vision-language prediction. The two tasks differ solely in what context the prediction conditions on. This is controlled by utilizing specific self-attention masks for the shared transformer network. To the best of our knowledge, VLP is the first reported model that achieves state-of-the-art results on both vision-language generation and understanding tasks, as disparate as image captioning and visual question answering, across three challenging benchmark datasets: COCO Captions, Flickr30k Captions, and VQA 2.0. The code and the pre-trained models are available at https://github.com/LuoweiZhou/VLP.
Video description is one of the most challenging problems in vision and language understanding due to the large variability both on the video and language side. Models, hence, typically shortcut the difficulty in recognition and generate plausible sentences that are based on priors but are not necessarily grounded in the video. In this work, we explicitly link the sentence to the evidence in the video by annotating each noun phrase in a sentence with the corresponding bounding box in one of the frames of a video. Our novel dataset, ActivityNet-Entities, is based on the challenging ActivityNet Captions dataset and augments it with 158k bounding box annotations, each grounding a noun phrase. This allows training video description models with this data, and importantly, evaluate how grounded or "true" such model are to the video they describe. To generate grounded captions, we propose a novel video description model which is able to exploit these bounding box annotations. We demonstrate the effectiveness of our model on our ActivityNet-Entities, but also show how it can be applied to image description on the Flickr30k Entities dataset. We achieve state-of-the-art performance on video description, video paragraph description, and image description and demonstrate our generated sentences are better grounded in the video.
Video action recognition, as a critical problem towards video understanding, has attracted increasing attention recently. To identify an action involving higher-order object interactions, we need to consider: 1) spatial relations among objects in a single frame; 2) temporal relations between different/same objects across multiple frames. However, previous approaches, e.g., 2D ConvNet + LSTM or 3D ConvNet, are either incapable of capturing relations between objects, or unable to handle streaming videos. In this paper, we propose a novel dynamic graph module to model object interactions in videos. We also devise two instantiations of our graph module: (i) visual graph, to capture visual similarity changes between objects; (ii) location graph, to capture relative location changes between objects. Distinct from previous models, the proposed graph module has the ability to process streaming videos in an aggressive manner. Combined with existing 3D action recognition ConvNets, our graph module can also boost ConvNets' performance, which demonstrates the flexibility of the module. We test our graph module on Something-Something dataset and achieve the state-of-the-art performance.