Visual Planning for Assistance (VPA) aims to predict a sequence of user actions required to achieve a specified goal based on a video showing the user’s progress. Although recent advances in multimodal large language models (MLLMs) have shown promising results in video understanding, long-horizon visual planning remains a challenging problem. We identify two challenges in training large MLLMs for video-based planning tasks: (1) scarcity of procedural annotations, limiting the model’s ability to learn procedural task dynamics effectively, and (2) inefficiency of next-token prediction objective to explicitly capture the structured action space for visual planning when compared to free-form, natural language. To tackle data scarcity, we introduce Auxiliary Task Augmentation. We design and train our model on auxiliary tasks relevant to long-horizon video-based planning (e.g., goal prediction) to augment the model’s planning ability. To more explicitly model the structured action space unique to visual planning tasks, we leverage Multi-token Prediction, extending traditional next-token prediction by using multiple heads to predict multiple future tokens during training. Our approach, VideoPlan, achieves state-of-the-art VPA performance on the COIN and CrossTask datasets, surpassing prior methods by 7.3% and 3.4%, respectively, when predicting 3 future actions. We further extend our method to the challenging Ego4D Long-term Action Anticipation task, and show that it is on par with the state-of-the-art approaches despite not using specialized egocentric features. Code is available at https://github.com/CeeZh/VideoPlan.

We investigate complex video question answering via chain-of-evidence reasoning — identifying sequences of temporal spans from multiple relevant parts of the video, together with visual evidence within them. Existing models struggle with multi-step reasoning as they uniformly sample a fixed number of frames, which can miss critical evidence distributed nonuniformly throughout the video. Moreover, they lack the ability to temporally localize such evidence in the broader context of the full video, which is required for answering complex questions. We propose a framework to enhance existing VideoQA datasets with evidence reasoning chains, automatically constructed by searching for optimal intervals of interest in the video with supporting evidence, that maximizes the likelihood of answering a given question. We train our model (VITED) to generate these evidence chains directly, enabling it to both localize evidence windows as well as perform multi-step reasoning across them in long-form video content. We show the value of our evidence-distilled models on a suite of long video QA benchmarks where we outperform state-of-the-art approaches that lack evidence reasoning capabilities.

We introduce ED-VTG, a method for fine-grained video temporal grounding utilizing multi-modal large language models. Our approach harnesses the capabilities of multimodal LLMs to jointly process text and video, in order to effectively localize natural language queries in videos through a two-stage process. Rather than being directly grounded, language queries are initially transformed into enriched sentences that incorporate missing details and cues to aid in grounding. In the second stage, these enriched queries are grounded, using a lightweight decoder, which specializes at predicting accurate boundaries conditioned on contextualized representations of the enriched queries. To mitigate noise and reduce the impact of hallucinations, our model is trained with a multiple-instance-learning objective that dynamically selects the optimal version of the query for each training sample. We demonstrate state-of-the-art results across various benchmarks in temporal video grounding and paragraph grounding settings. Experiments reveal that our method significantly outperforms all previously proposed LLM-based temporal grounding approaches and is either superior or comparable to specialized models, while maintaining a clear advantage against them in zero-shot evaluation scenarios.

Vision-language models are integral to computer vision research, yet many high-performing models remain closed-source, obscuring their data, design and training recipe. The research community has responded by using distillation from black-box models to label training data, achieving strong benchmark results, at the cost of measurable scientific progress. However, without knowing the details of the teacher model and its data sources, scientific progress remains difficult to measure. In this paper, we study building a Perception Language Model (PLM) in a fully open and reproducible framework for transparent research in image and video understanding. We analyze standard training pipelines without distillation from proprietary models and explore large-scale synthetic data to identify critical data gaps, particularly in detailed video understanding. To bridge these gaps, we release 2.8M human-labeled instances of fine-grained video question-answer pairs and spatio-temporally grounded video captions. Additionally, we introduce PLM–VideoBench, a suite for evaluating challenging video understanding tasks focusing on the ability to reason about ''what'', ''where'', ''when'', and ''how'' of a video. We make our work fully reproducible by providing data, training recipes, code & models.

We present Ego-Exo4D, a diverse, large-scale multimodal multiview video dataset and benchmark challenge. Ego-Exo4D centers around simultaneously-captured egocentric and exocentric video of skilled human activities (e.g., sports, music, dance, bike repair). 740 participants from 13 cities worldwide performed these activities in 123 different natural scene contexts, yielding long-form captures from 1 to 42 minutes each and 1,286 hours of video combined. The multimodal nature of the dataset is unprecedented: the video is accompanied by multichannel audio, eye gaze, 3D point clouds, camera poses, IMU, and multiple paired language descriptions—including a novel “expert commentary” done by coaches and teachers and tailored to the skilled-activity domain. To push the frontier of first-person video understanding of skilled human activity, we also present a suite of benchmark tasks and their annotations, including fine-grained activity understanding, proficiency estimation, cross-view translation, and 3D hand/body pose. All resources are open sourced to fuel new research in the community. https://ego-exo4d-data.org/

We present Ego-Exo4D, a diverse, large-scale multi-modal multiview video dataset and benchmark challenge. Ego-Exo4D centers around simultaneously-captured ego-centric and exocentric video of skilled human activities (e.g., sports, music, dance, bike repair). 740 participants from 13 cities worldwide performed these activities in 123 different natural scene contexts, yielding long-form captures from 1 to 42 minutes each and 1,286 hours of video combined. The multimodal nature of the dataset is un-precedented: the video is accompanied by multichannel audio, eye gaze, 3D point clouds, camera poses, IMU, and multiple paired language descriptions-including a novel “expert commentary” done by coaches and teachers and tailored to the skilled-activity domain. To push the frontier of first-person video understanding of skilled human activity, we also present a suite of benchmark tasks and their annotations, including fine-grained activity understanding, proficiency estimation, cross-view translation, and 3D hand/body pose. All resources are open sourced to fuel new research in the community.

Different video understanding tasks are typically treated in isolation, and even with distinct types of curated data (e.g., classifying sports in one dataset, tracking animals in another). However, in wearable cameras, the immersive egocentric perspective of a person engaging with the world around them presents an interconnected web of video understanding tasks—hand-object manipulations, navigation in the space, or human-human interactions—that unfold continuously, driven by the person's goals. We argue that this calls for a much more unified approach. We propose EgoTask Translation (EgoT2), which takes a collection of models optimized on separate tasks and learns to translate their outputs for improved performance on any or all of them at once. Unlike traditional transfer or multi-task learning, EgoT2's “flipped design” entails separate task-specific backbones and a task translator shared across all tasks, which captures synergies between even heterogeneous tasks and mitigates task competition. Demonstrating our model on a wide array of video tasks from Ego4D, we show its advantages over existing transfer paradigms and achieve top-ranked results on four of the Ego4D 2022 benchmark challenges.11Project webpage: https://vision.cs.utexas.edu/projects/egot2/.

Video-language pre-training (VLP) has become increasingly important due to its ability to generalize to various vision and language tasks. However, existing egocentric VLP frameworks utilize separate video and language encoders and learn task-specific cross-modal information only during fine-tuning, limiting the development of a unified system. In this work, we introduce the second generation of egocentric video-language pre-training (EgoVLPv2), a significant improvement from the previous generation, by incorporating cross-modal fusion directly into the video and language backbones. EgoVLPv2 learns strong video-text representation during pre-training and reuses the cross-modal attention modules to support different downstream tasks in a flexible and efficient manner, reducing fine-tuning costs. Moreover, our proposed fusion in the backbone strategy is more lightweight and compute-efficient than stacking additional fusion-specific layers. Extensive experiments on a wide range of VL tasks demonstrate the effectiveness of EgoVLPv2 by achieving consistent state-of-the-art performance over strong baselines across all downstream. Our project page can be found at https://shramanpramanick.github.io/EgoVLPv2/.

Human activities are goal-oriented and hierarchical, comprising primary goals at the top level, sequences of steps and substeps in the middle, and atomic actions at the lowest level. Recognizing human activities thus requires relating atomic actions and steps to their functional objectives (what the actions contribute to) and modeling their sequential and hierarchical dependencies towards achieving the goals. Current activity recognition research has primarily focused on only the lowest levels of this hierarchy, i.e., atomic or low-level actions, often in trimmed videos with annotations spanning only a few seconds. In this work, we introduce Ego4D Goal-Step, a new set of annotations on the recently released Ego4D with a novel hierarchical taxonomy of goal-oriented activity labels. It provides dense annotations for 48K procedural step segments (430 hours) and high-level goal annotations for 2,807 hours of Ego4D videos. Compared to existing procedural video datasets, it is substantially larger in size, contains hierarchical action labels (goals - steps - substeps), and provides goal-oriented auxiliary information including natural language summary description, step completion status, and step-to-goal relevance information. We take a data-driven approach to build our taxonomy, resulting in dense step annotations that do not suffer from poor label-data alignment issues resulting from a taxonomy defined a priori. Through comprehensive evaluations and analyses, we demonstrate how Ego4D Goal-Step supports exploring various questions in procedural activity understanding, including goal inference, step prediction, hierarchical relation learning, and long-term temporal modeling.

A critical object detection task is finetuning an existing model to detect novel objects, but the standard workflow requires bounding box annotations which are time-consuming and expensive to collect. Weakly supervised object detection (WSOD) offers an appealing alternative, where object detectors can be trained using image-level labels. However, the practical application of current WSOD models is limited, as they only operate at small data scales and require multiple rounds of training and refinement. To address this, we propose the Weakly Supervised Detection Transformer, which enables efficient knowledge transfer from a large-scale pretraining dataset to WSOD finetuning on hundreds of novel objects. Additionally, we leverage pretrained knowledge to improve the multiple instance learning (MIL) framework often used in WSOD methods. Our experiments show that our approach outperforms previous state-of-the-art models on large-scale novel object detection datasets, and our scaling study reveals that class quantity is more important than image quantity for WSOD pretraining.

Creating presentation materials requires complex multimodal reasoning skills to summarize key concepts and arrange them in a logical and visually pleasing manner. Can machines learn to emulate this laborious process? We present a novel task and approach for document-to-slide generation. Solving this involves document summarization, image and text retrieval, slide structure and layout prediction to arrange key elements in a form suitable for presentation. We propose a hierarchical sequence-to-sequence approach to tackle our task in an end-to-end manner. Our approach exploits the inherent structures within documents and slides and incorporates paraphrasing and layout prediction modules to generate slides. To help accelerate research in this domain, we release a dataset about 6K paired documents and slide decks used in our experiments. We show that our approach outperforms strong baselines and produces slides with rich content and aligned imagery.

Contrastive learning relies on an assumption that positive pairs contain related views that share certain underlying information about an instance, e.g., patches of an image or co-occurring multimodal signals of a video. What if this assumption is violated? The literature suggests that contrastive learning produces suboptimal representations in the presence of noisy views, e.g., false positive pairs with no apparent shared information. In this work, we pro-pose a new contrastive loss function that is robust against noisy views. We provide rigorous theoretical justifications by showing connections to robust symmetric losses for noisy binary classification and by establishing a new contrastive bound for mutual information maximization based on the Wasserstein distance measure. The proposed loss is completely modality-agnostic and a simple drop-in replacement for the InfoNCE loss, which makes it easy to apply to ex-isting contrastive frameworks. We show that our approach provides consistent improvements over the state-of-the-art on image, video, and graph contrastive learning bench-marks that exhibit a variety of real-world noise patterns.

We propose variational quasi-recurrent autoencoders (VQRAEs) to enable robust and efficient anomaly detection in time series in unsupervised settings. The proposed VQRAEs employs a judiciously designed objective function based on robust divergences, including a, ß, and, -divergence, making it possible to separate anomalies from normal data without the reliance on anomaly labels, thus achieving robustness and fully unsupervised training. To better capture temporal dependencies in time series data, VQRAEs are built upon quasi-recurrent neural networks, which employ convolution and gating mechanisms to avoid the inefficient recursive computations used by classic recurrent neural networks. Further, VQRAEs can be extended to bi-directional Bi VQRAEs that utilize bi-directional information to further improve the accuracy. The above design choices make VQRAEs not only robust and thus accurate, but also efficient at detecting anomalies in streaming settings. Experiments on five real-world time series offer insight into the design properties of VQRAEs and demonstrate that VQRAEs are capable of outperforming state-of-the-art methods.

Visual attention helps achieve robust perception under noise, corruption, and distribution shifts in human vision, which are areas where modern neural networks still fall short. We present VARS, Visual Attention from Recurrent Sparse reconstruction, a new attention formulation built on two prominent features of the human visual attention mechanism: recurrency and sparsity. Related features are grouped together via recurrent connections between neurons, with salient objects emerging via sparse regularization. VARS adopts an attractor network with recurrent connections that converges toward a stable pattern over time. Network layers are represented as ordinary differential equations (ODEs), formulating attention as a recurrent attractor network that equivalently optimizes the sparse reconstruction of input using a dictionary of “templates” encoding underlying patterns of data. We show that self-attention is a special case of VARS with a single-step optimization and no sparsity constraint. VARS can be readily used as a replacement for self-attention in popular vision transformers, consistently improving their robustness across various benchmarks.

Learning representations that generalize across tasks and domains is challenging yet necessary for autonomous systems. Although task-driven approaches are appealing, de-signing models specific to each application can be difficult in the face of limited data, especially when dealing with highly variable multimodal input spaces arising from different tasks in different environments. We introduce the first general-purpose pretraining pipeline, COntrastive Multimodal Pretraining for AutonomouS Systems (COMPASS), to overcome the limitations of task-specific models and existing pretraining approaches. COMPASS constructs a multimodal graph by considering the essential information for autonomous systems and the proper-ties of different modalities. Through this graph, multimodal signals are connected and mapped into two factorized spatio-temporal latent spaces: a “motion pattern space” and a “current state space.” By learning from multimodal correspondences in each latent space, COMPASS creates state representations that models necessary information such as temporal dynamics, geometry, and semantics. We pretrain COMPASS on a large-scale multimodal simulation dataset TartanAir [1] and evaluate it on drone navigation, vehicle racing, and visual odometry tasks. The experiments indicate that COMPASS can tackle all three scenarios and can also generalize to unseen environments and real-world data.11Our code implementation can be found at https://github.com/microsoft/COMPASS

The natural association between visual observations and their corresponding sound provides powerful self-supervisory signals for learning video representations, which makes the ever-growing amount of online videos an attractive source of training data. However, large portions of online videos contain irrelevant audio-visual signals because of edited/overdubbed audio, and models trained on such uncurated videos have shown to learn suboptimal representations. Therefore, existing self-supervised approaches rely on datasets with predetermined taxonomies of semantic concepts, where there is a high chance of audio-visual correspondence. Unfortunately, constructing such datasets require labor intensive manual annotation and/or verification, which severely limits the utility of online videos for large-scale learning. In this work, we present an automatic dataset curation approach based on subset optimization where the objective is to maximize the mutual information between audio and visual channels in videos. We demonstrate that our approach finds videos with high audio-visual correspondence and show that self-supervised models trained on our data achieve competitive performances compared to models trained on existing manually curated datasets. The most significant benefit of our approach is scalability: We release ACAV100M that contains 100 million videos with high audio-visual correspondence, ideal for self-supervised video representation learning.

Contrastive learning has delivered impressive results for various tasks in the self-supervised regime. However, existing approaches optimize for learning representations specific to downstream scenarios, i.e., global representations suitable for tasks such as classification or local representations for tasks such as detection and localization. While they produce satisfactory results in the intended downstream scenarios, they often fail to generalize to tasks that they were not originally designed for. In this work, we propose to learn video representations that generalize to both the tasks which require global semantic information (e.g., classification) and the tasks that require local fine-grained spatio-temporal information (e.g., localization). We achieve this by optimizing two contrastive objectives that together encourage our model to learn global-local visual information given audio signals. We show that the two objectives mutually improve the generalizability of the learned global-local representations, significantly outperforming their disjointly learned counterparts. We demonstrate our approach on various tasks including action/sound classification, lipreading, deepfake detection, event and sound localization.

Near-duplicate video retrieval (NDVR) is an important and challenging problem due to the increasing amount of videos uploaded to the Internet. In this paper, we propose an attention-based deep metric learning method for NDVR. Our method is based on well-established principles: We leverage two-stream networks to combine RGB and optical flow features, and incorporate an attention module to effectively deal with distractor frames commonly observed in near duplicate videos. We further aggregate the features corresponding to multiple video segments to enhance the discriminative power. The whole system is trained using a deep metric learning objective with a Siamese architecture. Our experiments show that the attention module helps eliminate redundant and noisy frames, while focusing on visually relevant frames for solving NVDR. We evaluate our approach on recent large-scale NDVR datasets, CC_WEB_VIDEO, VCDB, FIVR and SVD. To demonstrate the generalization ability of our approach, we report results in both within- and cross-dataset settings, and show that the proposed method significantly outperforms state-of-the-art approaches.

Current multi-reference style transfer models for Text-to-Speech (TTS) perform sub-optimally on disjoints datasets, where one dataset contains only a single style class for one of the style dimensions.These models generally fail to produce style transfer for the dimension that is underrepresented in the dataset.In this paper, we propose an adversarial cycle consistency training scheme with paired and unpaired triplets to ensure the use of information from all style dimensions.During training, we incorporate unpaired triplets with randomly selected reference audio samples and encourage the synthesized speech to preserve the appropriate styles using adversarial cycle consistency.We use this method to transfer emotion from a dataset containing four emotions to a dataset with only a single emotion.This results in a 78% improvement in style transfer (based on emotion classification) with minimal reduction in fidelity and naturalness.In subjective evaluations our method was consistently rated as closer to the reference style than the baseline.Synthesized speech samples are available at: https://sites.google.

Training deep neural networks typically requires large amounts of labeled data which may be scarce or expensive to obtain for a particular target domain. As an alternative, we can leverage webly-supervised data (i.e. results from a public search engine) which are relatively plentiful but may contain noisy results. In this work, we propose a novel two-stage approach to learn a video classifier using webly-supervised data. We argue that learning appearance features and temporal features sequentially, rather than jointly, is an easier optimization for this task. We show this by first learning an image model from web images, which is used to initialize and train a video model. Our model applies domain adaptation to account for potential domain shift present between the source domain (webly-supervised data) and target domain, and also accounts for noise by adding a novel attention component. We report results competitive with state-of-the-art for webly-supervised approaches (while simplifying the training process) on UCF-101 and also evaluate on Kinetics for comparison.

Visual-semantic embedding aims to find a shared latent space where related visual and textual instances are close to each other. Most current methods learn injective embedding functions that map an instance to a single point in the shared space. Unfortunately, injective embedding cannot effectively handle polysemous instances with multiple possible meanings; at best, it would find an average representation of different meanings. This hinders its use in real-world scenarios where individual instances and their cross-modal associations are often ambiguous. In this work, we introduce Polysemous Instance Embedding Networks (PIE-Nets) that compute multiple and diverse representations of an instance by combining global context with locally-guided features via multi-head self-attention and residual learning. To learn visual-semantic embedding, we tie-up two PIE-Nets and optimize them jointly in the multiple instance learning framework. Most existing work on cross-modal retrieval focus on image-text pairs of data. Here, we also tackle a more challenging case of video-text retrieval. To facilitate further research in video-text retrieval, we release a new dataset of 50K video-sentence pairs collected from social media, dubbed MRW (my reaction when). We demonstrate our approach on both image-text and video-text retrieval scenarios using MS-COCO, TGIF, and our new MRW dataset.

Deep generative models have led to significant advances in cross-modal generation such as text-to-image synthesis. Training these models typically requires paired data with direct correspondence between modalities. We introduce the novel problem of translating instances from one modality to another without paired data by leveraging an intermediate modality shared by the two other modalities. To demonstrate this, we take the problem of translating images to speech. In this case, one could leverage disjoint datasets with one shared modality, e.g., image-text pairs and text-speech pairs, with text as the shared modality. We call this problem “skip-modal generation” because the shared modality is skipped during the generation process. We propose a multimodal information bottleneck approach that learns the correspondence between modalities from unpaired data (image and speech) by leveraging the shared modality (text). We address fundamental challenges of skip-modal generation: 1) learning multimodal representations using a single model, 2) bridging the domain gap between two unrelated datasets, and 3) learning the correspondence between modalities from unpaired data. We show qualitative results on image-to-speech synthesis; this is the first time such results have been reported in the literature. We also show that our approach improves performance on traditional cross-modal generation, suggesting that it improves data efficiency in solving individual tasks.

Models that are learned from real-world data are often biased because the data used to train them is biased. This can propagate systemic human biases that exist and ultimately lead to inequitable treatment of people, especially minorities. To characterize bias in learned classifiers, existing approaches rely on human oracles labeling real-world examples to identify the "blind spots" of the classifiers; these are ultimately limited due to the human labor required and the finite nature of existing image examples. We propose a simulation-based approach for interrogating classifiers using generative adversarial models in a systematic manner. We incorporate a progressive conditional generative model for synthesizing photo-realistic facial images and Bayesian Optimization for an efficient interrogation of independent facial image classification systems. We show how this approach can be used to efficiently characterize racial and gender biases in commercial systems.

Video prediction aims to generate realistic future frames by learning dynamic visual patterns. One fundamental challenge is to deal with future uncertainty: How should a model behave when there are multiple correct, equally probable future? We propose an Appearance-Motion Conditional GAN to address this challenge. We provide appearance and motion information as conditions that specify how the future may look like, reducing the level of uncertainty. Our model consists of a generator, two discriminators taking charge of appearance and motion pathways, and a perceptual ranking module that encourages videos of similar conditions to look similar. To train our model, we develop a novel conditioning scheme that consists of different combinations of appearance and motion conditions. We evaluate our model using facial expression and human action datasets and report favorable results compared to existing methods.

Given a still photograph, one can imagine how dynamic objects might move against a static background. This idea has been actualized in the form of cinemagraphs, where the motion of particular objects within a still image is repeated, giving the viewer a sense of animation. In this paper, we learn computational models that can generate cinemagraph sequences automatically given a single image. To generate cinemagraphs, we explore combining generative models with a recurrent neural network and deep Q-networks to enhance the power of sequence generation. To enable and evaluate these models we make use of two datasets, one synthetically generated and the other containing real video generated cinemagraphs. Both qualitative and quantitative evaluations demonstrate the effectiveness of our models on the synthetic and real datasets.

Vision and language understanding has emerged as a subject undergoing intense study in Artificial Intelligence. Among many tasks in this line of research, visual question answering (VQA) has been one of the most successful ones, where the goal is to learn a model that understands visual content at region-level details and finds their associations with pairs of questions and answers in the natural language form. Despite the rapid progress in the past few years, most existing work in VQA have focused primarily on images. In this paper, we focus on extending VQA to the video domain and contribute to the literature in three important ways. First, we propose three new tasks designed specifically for video VQA, which require spatio-temporal reasoning from videos to answer questions correctly. Next, we introduce a new large-scale dataset for video VQA named TGIF-QA that extends existing VQA work with our new tasks. Finally, we propose a dual-LSTM based approach with both spatial and temporal attention, and show its effectiveness over conventional VQA techniques through empirical evaluations.

Learning to rank has recently emerged as an attractive technique to train deep convolutional neural networks for various computer vision tasks. Pairwise ranking, in particular, has been successful in multi-label image classification, achieving state-of-the-art results on various benchmarks. However, most existing approaches use the hinge loss to train their models, which is non-smooth and thus is difficult to optimize especially with deep networks. Furthermore, they employ simple heuristics, such as top-k or thresholding, to determine which labels to include in the output from a ranked list of labels, which limits their use in the real-world setting. In this work, we propose two techniques to improve pairwise ranking based multi-label image classification by solving the aforementioned problems: (1) we propose a novel loss function for pairwise ranking, which is smooth everywhere, and (2) we incorporate a label decision module into the model, estimating the optimal confidence thresholds for each visual concept. We provide theoretical analyses of our loss function from the point of view of the Bayes consistency and risk minimization, and show its benefit over existing pairwise ranking formulations. We also demonstrate the effectiveness of our approach on two large-scale datasets, NUS-WIDE and MS-COCO, achieving the best reported result in the literature.

The ability of learning from noisy labels is very useful in many visual recognition tasks, as a vast amount of data with noisy labels are relatively easy to obtain. Traditionally, label noise has been treated as statistical outliers, and techniques such as importance re-weighting and bootstrapping have been proposed to alleviate the problem. According to our observation, the real-world noisy labels exhibit multimode characteristics as the true labels, rather than behaving like independent random outliers. In this work, we propose a unified distillation framework to use “side” information, including a small clean dataset and label relations in knowledge graph, to “hedge the risk” of learning from noisy labels. Unlike the traditional approaches evaluated based on simulated label noises, we propose a suite of new benchmark datasets, in Sports, Species and Artifacts domains, to evaluate the task of learning from noisy labels in the practical setting. The empirical study demonstrates the effectiveness of our proposed method in all the domains.

Video consumption is being shifted from sit-and-watch to selective skimming. Existing video player interfaces, however, only provide indirect manipulation to support this emerging behavior. Video summarization alleviates this issue to some extent, shortening a video based on the desired length of a summary as an input variable. But an optimal length of a summarized video is often not available in advance. Moreover, the user cannot edit the summary once it is produced, limiting its practical applications. We argue that video summarization should be an interactive, mixed-initiative process in which users have control over the summarization procedure while algorithms help users achieve their goal via video understanding. In this paper, we introduce ElasticPlay, a mixed-initiative approach that combines an advanced video summarization technique with direct interface manipulation to help users control the video summarization process. Users can specify a time budget for the remaining content while watching a video; our system then immediately updates the playback plan using our proposed cut-and-forward algorithm, determining which parts to skip or to fast-forward. This interactive process allows users to fine-tune the summarization result with immediate feedback. We show that our system outperforms existing video summarization techniques on the TVSum50 dataset. We also report two lab studies (22 participants) and a Mechanical Turk deployment study (60 participants), and show that the participants responded favorably to ElasticPlay.

Animated GIFs have been around since 1987 and recently gained more popularity on social networking sites. Tumblr, a large social networking and micro blogging platform, is a popular venue to share animated GIFs. Tumblr users follow blogs, generating a feed or posts, and choose to "like' or to "reblog' favored posts. In this paper, we use these actions as signals to analyze the engagement of over 3.9 million posts, and conclude that animated GIFs are significantly more engaging than other kinds of media. We follow this finding with deeper visual analysis of nearly 100k animated GIFs and pair our results with interviews with 13 Tumblr users to find out what makes animated GIFs engaging. We found that the animation, lack of sound, immediacy of consumption, low bandwidth and minimal time demands, the storytelling capabilities and utility for expressing emotions were significant factors in making GIFs the most engaging content on Tumblr. We also found that engaging GIFs contained faces and had higher motion energy, uniformity, resolution and frame rate. Our findings connect to media theories and have implications in design of effective content dashboards, video summarization tools and ranking algorithms to enhance engagement.

Thumbnails play such an important role in online videos. As the most representative snapshot, they capture the essence of a video and provide the first impression to the viewers; ultimately, a great thumbnail makes a video more attractive to click and watch. We present an automatic thumbnail selection system that exploits two important characteristics commonly associated with meaningful and attractive thumbnails: high relevance to video content and superior visual aesthetic quality. Our system selects attractive thumbnails by analyzing various visual quality and aesthetic metrics of video frames, and performs a clustering analysis to determine the relevance to video content, thus making the resulting thumbnails more representative of the video. On the task of predicting thumbnails chosen by professional video editors, we demonstrate the effectiveness of our system against six baseline methods, using a real-world dataset of 1,118 videos collected from Yahoo Screen. In addition, we study what makes a frame a good thumbnail by analyzing the statistical relationship between thumbnail frames and non-thumbnail frames in terms of various image quality features. Our study suggests that the selection of a good thumbnail is highly correlated with objective visual quality metrics, such as the frame texture and sharpness, implying the possibility of building an automatic thumbnail selection system based on visual aesthetics.

We introduce the novel problem of automatically generating animated GIFs from video. GIFs are short looping video with no sound, and a perfect combination between image and video that really capture our attention. GIFs tell a story, express emotion, turn events into humorous moments, and are the new wave of photojournalism. We pose the question: Can we automate the entirely manual and elaborate process of GIF creation by leveraging the plethora of user generated GIF content? We propose a Robust Deep RankNet that, given a video, generates a ranked list of its segments according to their suitability as GIF. We train our model to learn what visual content is often selected for GIFs by using over 100K user generated GIFs and their corresponding video sources. We effectively deal with the noisy web data by proposing a novel adaptive Huber loss in the ranking formulation. We show that our approach is robust to outliers and picks up several patterns that are frequently present in popular animated GIFs. On our new large-scale benchmark dataset, we show the advantage of our approach over several state-of-the-art methods.

With the recent popularity of animated GIFs on social media, there is need for ways to index them with rich meta-data. To advance research on animated GIF understanding, we collected a new dataset, Tumblr GIF (TGIF), with 100K animated GIFs from Tumblr and 120K natural language descriptions obtained via crowdsourcing. The motivation for this work is to develop a testbed for image sequence description systems, where the task is to generate natural language descriptions for animated GIFs or video clips. To ensure a high quality dataset, we developed a series of novel quality controls to validate free-form text input from crowd-workers. We show that there is unambiguous association between visual content and natural language descriptions in our dataset, making it an ideal benchmark for the visual content captioning task. We perform extensive statistical analyses to compare our dataset to existing image and video description datasets. Next, we provide baseline results on the animated GIF description task, using three representative techniques: nearest neighbor, statistical machine translation, and recurrent neural networks. Finally, we show that models fine-tuned from our animated GIF description dataset can be helpful for automatic movie description.

Automatic detection of interesting moments in video has many real-world applications such as video summarization and efficient online video browsing. In this paper, we present a lightweight and scalable solution to this problem based on user mouse activity while watching video. Unlike previous approaches that analyze video content to infer the interestingness, we leverage the implicit user feedback obtained from thousands of online video watching sessions. This makes our method computationally efficient and scalable to billions of videos. Most importantly, our approach can handle a variety of video genres because we make no assumption on what constitutes interestingness: we let the crowd tell us through their mouse activity. By analyzing 106,212 user sessions collected from a popular online video website, we show that mouse activity is highly indicative of interestingness, and that our approach has competitive performance to several state-of-the-art methods.

We present video co-summarization, a novel perspective to video summarization that exploits visual co-occurrence across multiple videos. Motivated by the observation that important visual concepts tend to appear repeatedly across videos of the same topic, we propose to summarize a video by finding shots that co-occur most frequently across videos collected using a topic keyword. The main technical challenge is dealing with the sparsity of co-occurring patterns, out of hundreds to possibly thousands of irrelevant shots in videos being considered. To deal with this challenge, we developed a Maximal Biclique Finding (MBF) algorithm that is optimized to find sparsely co-occurring patterns, discarding less co-occurring patterns even if they are dominant in one video. Our algorithm is parallelizable with closed-form updates, thus can easily scale up to handle a large number of videos simultaneously. We demonstrate the effectiveness of our approach on motion capture and self-compiled YouTube datasets. Our results suggest that summaries generated by visual co-occurrence tend to match more closely with human generated summaries, when compared to several popular unsupervised techniques.

Video summarization is a challenging problem in part because knowing which part of a video is important requires prior knowledge about its main topic. We present TVSum, an unsupervised video summarization framework that uses title-based image search results to find visually important shots. We observe that a video title is often carefully chosen to be maximally descriptive of its main topic, and hence images related to the title can serve as a proxy for important visual concepts of the main topic. However, because titles are free-formed, unconstrained, and often written ambiguously, images searched using the title can contain noise (images irrelevant to video content) and variance (images of different topics). To deal with this challenge, we developed a novel co-archetypal analysis technique that learns canonical visual concepts shared between video and images, but not in either alone, by finding a joint-factorial representation of two data sets. We introduce a new benchmark dataset, TVSum50, that contains 50 videos and their shot-level importance scores annotated via crowdsourcing. Experimental results on two datasets, SumMe and TVSum50, suggest our approach produces superior quality summaries compared to several recently proposed approaches.

We present FluxFlow, an interactive visual analysis system for revealing and analyzing anomalous information spreading in social media. Everyday, millions of messages are created, commented, and shared by people on social media websites, such as Twitter and Facebook. This provides valuable data for researchers and practitioners in many application domains, such as marketing, to inform decision-making. Distilling valuable social signals from the huge crowd's messages, however, is challenging, due to the heterogeneous and dynamic crowd behaviors. The challenge is rooted in data analysts' capability of discerning the anomalous information behaviors, such as the spreading of rumors or misinformation, from the rest that are more conventional patterns, such as popular topics and newsworthy events, in a timely fashion. FluxFlow incorporates advanced machine learning algorithms to detect anomalies, and offers a set of novel visualization designs for presenting the detected threads for deeper analysis. We evaluated FluxFlow with real datasets containing the Twitter feeds captured during significant events such as Hurricane Sandy. Through quantitative measurements of the algorithmic performance and qualitative interviews with domain experts, the results show that the back-end anomaly detection model is effective in identifying anomalous retweeting threads, and its front-end interactive visualizations are intuitive and useful for analysts to discover insights in data and comprehend the underlying analytical model.

Recent progress has shown that learning from hierarchical feature representations leads to improvements in various computer vision tasks. Motivated by the observation that human activity data contains information at various temporal resolutions, we present a hierarchical sequence summarization approach for action recognition that learns multiple layers of discriminative feature representations at different temporal granularities. We build up a hierarchy dynamically and recursively by alternating sequence learning and sequence summarization. For sequence learning we use CRFs with latent variables to learn hidden spatio-temporal dynamics, for sequence summarization we group observations that have similar semantic meaning in the latent space. For each layer we learn an abstract feature representation through non-linear gate functions. This procedure is repeated to obtain a hierarchical sequence summary representation. We develop an efficient learning method to train our model and show that its complexity grows sub linearly with the size of the hierarchy. Experimental results show the effectiveness of our approach, achieving the best published results on the Arm Gesture and Canal9 datasets.

Obtaining a compact and discriminative representation of facial and body expressions is a difficult problem in emotion recognition. Part of the difficulty is capturing microexpressions, i.e., short, involuntary expressions that last for only a fraction of a second: at a micro-temporal scale, there are so many other subtle face and body movements that do not convey semantically meaningful information. We present a novel approach to this problem by exploiting the sparsity of the frequent micro-temporal motion patterns. Local space-time features are extracted over the face and body region for a very short time period, e.g., few milliseconds. A codebook of microexpressions is learned from the data and used to encode the features in a sparse manner. This allows us to obtain a representation that captures the most salient motion patterns of the face and body at a micro-temporal scale. Experiments performed on the AVEC 2012 dataset show our approach achieving the best published performance on the arousal dimension based solely on visual features. We also report experimental results on audio-visual emotion recognition, comparing early and late data fusion techniques.

Many human action recognition tasks involve data that can be factorized into multiple views such as body postures and hand shapes. These views often interact with each other over time, providing important cues to understanding the action. We present multi-view latent variable discriminative models that jointly learn both view-shared and view-specific sub-structures to capture the interaction between views. Knowledge about the underlying structure of the data is formulated as a multi-chain structured latent conditional model, explicitly learning the interaction between multiple views using disjoint sets of hidden variables in a discriminative manner. The chains are tied using a predetermined topology that repeats over time. We present three topologies - linked, coupled, and linked-coupled - that differ in the type of interaction between views that they model. We evaluate our approach on both segmented and unsegmented human action recognition tasks, using the ArmGesture, the NATOPS, and the ArmGesture-Continuous data. Experimental results show that our approach outperforms previous state-of-the-art action recognition models.

Multimodal human behavior analysis is a challenging task due to the presence of complex nonlinear correlations and interactions across modalities. We present a novel approach to this problem based on Kernel Canonical Correlation Analysis (KCCA) and Multi-view Hidden Conditional Random Fields (MV-HCRF). Our approach uses a nonlinear kernel to map multimodal data to a high-dimensional feature space and finds a new projection of the data that maximizes the correlation across modalities. We use a multi-chain structured graphical model with disjoint sets of latent variables, one set per modality, to jointly learn both view-shared and view-specific sub-structures of the projected data, capturing interaction across modalities explicitly. We evaluate our approach on a task of agreement and disagreement recognition from nonverbal audio-visual cues using the Canal 9 dataset. Experimental results show that KCCA makes capturing nonlinear hidden dynamics easier and MV-HCRF helps learning interaction across modalities.

Intelligent gesture recognition systems open a new era of natural human-computer interaction: Gesturing is instinctive and a skill we all have, so it requires little or no thought, leaving the focus on the task itself, as it should be, not on the interaction modality. We present a new approach to gesture recognition that attends to both body and hands, and interprets gestures continuously from an unsegmented and unbounded input stream. This article describes the whole procedure of continuous body and hand gesture recognition, from the signal acquisition to processing, to the interpretation of the processed signals. Our system takes a vision-based approach, tracking body and hands using a single stereo camera. Body postures are reconstructed in 3D space using a generative model-based approach with a particle filter, combining both static and dynamic attributes of motion as the input feature to make tracking robust to self-occlusion. The reconstructed body postures guide searching for hands. Hand shapes are classified into one of several canonical hand shapes using an appearance-based approach with a multiclass support vector machine. Finally, the extracted body and hand features are combined and used as the input feature for gesture recognition. We consider our task as an online sequence labeling and segmentation problem. A latent-dynamic conditional random field is used with a temporal sliding window to perform the task continuously. We augment this with a novel technique called multilayered filtering, which performs filtering both on the input layer and the prediction layer. Filtering on the input layer allows capturing long-range temporal dependencies and reducing input signal noise; filtering on the prediction layer allows taking weighted votes of multiple overlapping prediction results as well as reducing estimation noise. We tested our system in a scenario of real-world gestural interaction using the NATOPS dataset, an official vocabulary of aircraft handling gestures. Our experimental results show that: (1) the use of both static and dynamic attributes of motion in body tracking allows statistically significant improvement of the recognition performance over using static attributes of motion alone; and (2) the multilayered filtering statistically significantly improves recognition performance over the nonfiltering method. We also show that, on a set of twenty-four NATOPS gestures, our system achieves a recognition accuracy of 75.37%.

We present a new approach to multi-signal gesture recognition that attends to simultaneous body and hand movements. The system examines temporal sequences of dual-channel input signals obtained via statistical inference that indicate 3D body pose and hand pose. Learning gesture patterns from these signals can be quite challenging due to the existence of long-range temporal-dependencies and low signal-to-noise ratio (SNR). We incorporate a Gaussian temporal-smoothing kernel into the inference framework, capturing long-range temporal-dependencies and increasing the SNR efficiently. An extensive set of experiments was performed, allowing us to (1) show that combining body and hand signals significantly improves the recognition accuracy; (2) report on which features of body and hands are most informative; and (3) show that using a Gaussian temporal-smoothing significantly improves gesture recognition accuracy.

We present a unified framework for body and hand tracking, the output of which can be used for understanding simultaneously performed body-and-hand gestures. The framework uses a stereo camera to collect 3D images, and tracks body and hand together, combining various existing techniques to make tracking tasks efficient. In addition, we introduce a multi-signal gesture database: the NATOPS aircraft handling signals. Unlike previous gesture databases, this data requires knowledge about both body and hand in order to distinguish gestures. It is also focused on a clearly defined gesture vocabulary from a real-world scenario that has been refined over many years. The database includes 24 body-and-hand gestures, and provides both gesture video clips and the body and hand features we extracted.