We present a novel Cross-Class Relevance Learning approach for the task of temporal concept localization. Most localization architectures rely on feature extraction layers followed by a classification layer which outputs class probabilities for each segment. However, in many real-world applications classes can exhibit complex relationships that are difficult to model with this architecture. In contrast, we propose to incorporate target class and class-related features as input, and learn a pairwise binary model to predict general segment to class relevance. This facilitates learning of shared information between classes, and allows for arbitrary class-specific feature engineering. We apply this approach to the 3rd YouTube-8M Video Understanding Challenge together with other leading models, and achieve first place out of over 280 teams. In this paper we describe our approach and show some empirical results.
Video understanding usually requires expensive computation that prohibits its deployment, yet videos contain significant spatiotemporal redundancy that can be exploited. In particular, operating directly on the motion vectors and residuals in the compressed video domain can significantly accelerate the compute, by not using the raw videos which demand colossal storage capacity. Existing methods approach this task as a multiple modalities problem. In this paper we are approaching the task in a completely different way; we are looking at the data from the compressed stream as a one unit clip and propose that the residual frames can replace the original RGB frames from the raw domain. Furthermore, we are using teacher-student method to aid the network in the compressed domain to mimic the teacher network in the raw domain. We show experiments on three leading datasets (HMDB51, UCF1, and Kinetics) that approach state-of-the-art accuracy on raw video data by using compressed data. Our model MFCD-Net outperforms prior methods in the compressed domain and more importantly, our model has 11X fewer parameters and 3X fewer Flops, dramatically improving the efficiency of video recognition inference. This approach enables applying neural networks exclusively in the compressed domain without compromising accuracy while accelerating performance.
Temporal localization remains an important challenge in video understanding. In this work, we present our solution to the 3rd YouTube-8M Video Understanding Challenge organized by Google Research. Participants were required to build a segment-level classifier using a large-scale training data set with noisy video-level labels and a relatively small-scale validation data set with accurate segment-level labels. We formulated the problem as a multiple instance multi-label learning and developed an attention-based mechanism to selectively emphasize the important frames by attention weights. The model performance is further improved by constructing multiple sets of attention networks. We further fine-tuned the model using the segment-level data set. Our final model consists of an ensemble of attention/multi-attention networks, deep bag of frames models, recurrent neural networks and convolutional neural networks. It ranked 13th on the private leader board and stands out for its efficient usage of resources.
Videos capture events that typically contain multiple sequential, and simultaneous, actions even in the span of only a few seconds. However, most large-scale datasets built to train models for action recognition in video only provide a single label per video. Consequently, models can be incorrectly penalized for classifying actions that exist in the videos but are not explicitly labeled and do not learn the full spectrum of information present in each video in training. Towards this goal, we present the Multi-Moments in Time dataset (M-MiT) which includes over two million action labels for over one million three second videos. This multi-label dataset introduces novel challenges on how to train and analyze models for multi-action detection. Here, we present baseline results for multi-action recognition using loss functions adapted for long tail multi-label learning, provide improved methods for visualizing and interpreting models trained for multi-label action detection and show the strength of transferring models trained on M-MiT to smaller datasets.
Video summarization aims to extract keyframes/shots from a long video. Previous methods mainly take diversity and representativeness of generated summaries as prior knowledge in algorithm design. In this paper, we formulate video summarization as a content-based recommender problem, which should distill the most useful content from a long video for users who suffer from information overload. A scalable deep neural network is proposed on predicting if one video segment is a useful segment for users by explicitly modelling both segment and video. Moreover, we accomplish scene and action recognition in untrimmed videos in order to find more correlations among different aspects of video understanding tasks. Also, our paper will discuss the effect of audio and visual features in summarization task. We also extend our work by data augmentation and multi-task learning for preventing the model from early-stage overfitting. The final results of our model win the first place in ICCV 2019 CoView Workshop Challenge Track.
In this paper, we present and discuss a deep mixture model with online knowledge distillation (MOD) for large-scale video temporal concept localization, which is ranked 3rd in the 3rd YouTube-8M Video Understanding Challenge. Specifically, we find that by enabling knowledge sharing with online distillation, fintuning a mixture model on a smaller dataset can achieve better evaluation performance. Based on this observation, in our final solution, we trained and fintuned 12 NeXtVLAD models in parallel with a 2-layer online distillation structure. The experimental results show that the proposed distillation structure can effectively avoid overfitting and shows superior generalization performance. The code is publicly available at: https://github.com/linrongc/solution_youtube8m_v3
We propose a novel video understanding task by fusing knowledge-based and video question answering. First, we introduce KnowIT VQA, a video dataset with 24,282 human-generated question-answer pairs about a popular sitcom. The dataset combines visual, textual and temporal coherence reasoning together with knowledge-based questions, which need of the experience obtained from the viewing of the series to be answered. Second, we propose a video understanding model by combining the visual and textual video content with specific knowledge about the show. Our main findings are: (i) the incorporation of knowledge produces outstanding improvements for VQA in video, and (ii) the performance on KnowIT VQA still lags well behind human accuracy, indicating its usefulness for studying current video modelling limitations.
Temporal action detection is a fundamental yet challenging task in video understanding. Many of the state-of-the-art methods predict the boundaries of action instances based on predetermined anchors akin to the two-dimensional object detection detectors. However, it is hard to detect all the action instances with predetermined temporal scales because the durations of instances in untrimmed videos can vary from few seconds to several minutes. In this paper, we propose a novel action detection architecture named anchor-free one-stage temporal action detector (AFO-TAD). AFO-TAD achieves better performance for detecting action instances with arbitrary lengths and high temporal resolution, which can be attributed to two aspects. First, we design a receptive field adaption module which dynamically adjusts the receptive field for precise action detection. Second, AFO-TAD directly predicts the categories and boundaries at every temporal locations without predetermined anchors. Extensive experiments show that AFO-TAD improves the state-of-the-art performance on THUMOS'14.
Video understanding is a challenging problem with great impact on the abilities of autonomous agents working in the real-world. Yet, solutions so far have been computationally intensive, with the fastest algorithms running for more than half a second per video snippet on powerful GPUs. We propose a novel idea on video architecture learning - Tiny Video Networks - which automatically designs highly efficient models for video understanding. The tiny video models run with competitive performance for as low as 37 milliseconds per video on a CPU and 10 milliseconds on a standard GPU.
The automatic detection and tracking of general objects (like persons, animals or cars), text and logos in a video is crucial for many video understanding tasks, and usually real-time processing as required. We propose OmniTrack, an efficient and robust algorithm which is able to automatically detect and track objects, text as well as brand logos in real-time. It combines a powerful deep learning based object detector (YoloV3) with high-quality optical flow methods. Based on the reference YoloV3 C++ implementation, we did some important performance optimizations which will be described. The major steps in the training procedure for the combined detector for text and logo will be presented. We will describe then the OmniTrack algorithm, consisting of the phases preprocessing, feature calculation, prediction, matching and update. Several performance optimizations have been implemented there as well, like doing the object detection and optical flow calculation asynchronously. Experiments show that the proposed algorithm runs in real-time for standard definition ($720x576$) video on a PC with a Quadro RTX 5000 GPU.
Computer vision has undergone a dramatic revolution in performance, driven in large part through deep features trained on large-scale supervised datasets. However, much of these improvements have focused on static image analysis; video understanding has seen rather modest improvements. Even though new datasets and spatiotemporal models have been proposed, simple frame-by-frame classification methods often still remain competitive. We posit that current video datasets are plagued with implicit biases over scene and object structure that can dwarf variations in temporal structure. In this work, we build a video dataset with fully observable and controllable object and scene bias, and which truly requires spatiotemporal understanding in order to be solved. Our dataset, named CATER, is rendered synthetically using a library of standard 3D objects, and tests the ability to recognize compositions of object movements that require long-term reasoning. In addition to being a challenging dataset, CATER also provides a plethora of diagnostic tools to analyze modern spatiotemporal video architectures by being completely observable and controllable. Using CATER, we provide insights into some of the most recent state of the art deep video architectures.
This work presents the Video Platform for PyTorch (ViP), a deep learning-based framework designed to handle and extend to any problem domain based on videos. ViP supports (1) a single unified interface applicable to all video problem domains, (2) quick prototyping of video models, (3) executing large-batch operations with reduced memory consumption, and (4) easy and reproducible experimental setups. ViP's core functionality is built with flexibility and modularity in mind to allow for smooth data flow between different parts of the platform and benchmarking against existing methods. In providing a software platform that supports multiple video-based problem domains, we allow for more cross-pollination of models, ideas and stronger generalization in the video understanding research community.
Recent years have seen a dramatic expansion of tasks and datasets posed as question answering, from reading comprehension, semantic role labeling, and even machine translation, to image and video understanding. With this expansion, there are many differing views on the utility and definition of "question answering" itself. Some argue that its scope should be narrow, or broad, or that it is overused in datasets today. In this opinion piece, we argue that question answering should be considered a format which is sometimes useful for studying particular phenomena, not a phenomenon or task in itself. We discuss when a task is correctly described as question answering, and when a task is usefully posed as question answering, instead of using some other format.
Zero-Shot Action Recognition has attracted attention in the last years and many approaches have been proposed for recognition of objects, events and actions in images and videos. There is a demand for methods that can classify instances from classes that are not present in the training of models, especially in the complex problem of automatic video understanding, since collecting, annotating and labeling videos are difficult and laborious tasks. We have identified that there are many methods available in the literature, however, it is difficult to categorize which techniques can be considered state of the art. Despite the existence of some surveys about zero-shot action recognition in still images and experimental protocol, there is no work focused on videos. Therefore, we present a survey of the methods that comprise techniques to perform visual feature extraction and semantic feature extraction as well to learn the mapping between these features considering specifically zero-shot action recognition in videos. We also provide a complete description of datasets, experiments and protocols, presenting open issues and directions for future work, essential for the development of the computer vision research field.
Temporally localizing actions in a video is a fundamental challenge in video understanding. Most existing approaches have often drawn inspiration from image object detection and extended the advances, e.g., SSD and Faster R-CNN, to produce temporal locations of an action in a 1D sequence. Nevertheless, the results can suffer from robustness problem due to the design of predetermined temporal scales, which overlooks the temporal structure of an action and limits the utility on detecting actions with complex variations. In this paper, we propose to address the problem by introducing Gaussian kernels to dynamically optimize temporal scale of each action proposal. Specifically, we present Gaussian Temporal Awareness Networks (GTAN) --- a new architecture that novelly integrates the exploitation of temporal structure into an one-stage action localization framework. Technically, GTAN models the temporal structure through learning a set of Gaussian kernels, each for a cell in the feature maps. Each Gaussian kernel corresponds to a particular interval of an action proposal and a mixture of Gaussian kernels could further characterize action proposals with various length. Moreover, the values in each Gaussian curve reflect the contextual contributions to the localization of an action proposal. Extensive experiments are conducted on both THUMOS14 and ActivityNet v1.3 datasets, and superior results are reported when comparing to state-of-the-art approaches. More remarkably, GTAN achieves 1.9% and 1.1% improvements in mAP on testing set of the two datasets.
For video understanding, namely analyzing who did what in a video, actions along with objects are primary elements. Most studies on actions have handled recognition problems for a well‐trimmed video and focused on enhancing their classification performance. However, action detection, including localization as well as recognition, is required because, in general, actions intersect in time and space. In addition, most studies have not considered extensibility for a newly added action that has been previously trained. Therefore, proposed in this paper is an extensible hierarchical method for detecting generic actions, which combine object movements and spatial relations between two objects, and inherited actions, which are determined by the related objects through an ontology and rule based methodology. The hierarchical design of the method enables it to detect any interactive actions based on the spatial relations between two objects. The method using object information achieves an F‐measure of 90.27%. Moreover, this paper describes the extensibility of the method for a new action contained in a video from a video domain that is different from the dataset used.
Understanding temporal information and how the visual world changes over time is a fundamental ability of intelligent systems. In video understanding, temporal information is at the core of many current challenges, including compression, efficient inference, motion estimation or summarization. However, in current video datasets it has been observed that action classes can often be recognized without any temporal information from a single frame of video. As a result, both benchmarking and training in these datasets may give an unintentional advantage to models with strong image understanding capabilities, as opposed to those with strong temporal understanding. In this paper we address this problem head on by identifying action classes where temporal information is actually necessary to recognize them and call these "temporal classes". Selecting temporal classes using a computational method would bias the process. Instead, we propose a methodology based on a simple and effective human annotation experiment. We remove just the temporal information by shuffling frames in time and measure if the action can still be recognized. Classes that cannot be recognized when frames are not in order are included in the temporal Dataset. We observe that this set is statistically different from other static classes, and that performance in it correlates with a network's ability to capture temporal information. Thus we use it as a benchmark on current popular networks, which reveals a series of interesting facts. We also explore the effect of training on the temporal dataset, and observe that this leads to better generalization in unseen classes, demonstrating the need for more temporal data. We hope that the proposed dataset of temporal categories will help guide future research in temporal modeling for better video understanding.
Action localization in untrimmed videos is an important topic in the field of video understanding. However, existing action localization methods are restricted to a pre-defined set of actions and cannot localize unseen activities. Thus, we consider a new task to localize unseen activities in videos via image queries, named Image-Based Activity Localization. This task faces three inherent challenges: (1) how to eliminate the influence of semantically inessential contents in image queries; (2) how to deal with the fuzzy localization of inaccurate image queries; (3) how to determine the precise boundaries of target segments. We then propose a novel self-attention interaction localizer to retrieve unseen activities in an end-to-end fashion. Specifically, we first devise a region self-attention method with relative position encoding to learn fine-grained image region representations. Then, we employ a local transformer encoder to build multi-step fusion and reasoning of image and video contents. We next adopt an order-sensitive localizer to directly retrieve the target segment. Furthermore, we construct a new dataset ActivityIBAL by reorganizing the ActivityNet dataset. The extensive experiments show the effectiveness of our method.
Although a video is effectively a sequence of images, visual perception systems typically model images and videos separately, thus failing to exploit the correlation and the synergy provided by these two media. While a few prior research efforts have explored the benefits of leveraging still-image datasets for video analysis, or vice-versa, most of these attempts have been limited to pretraining a model on one type of visual modality and then adapting it via finetuning on the other modality. In contrast, in this paper we introduce a framework that enables joint training of a unified model on mixed collections of image and video examples spanning different tasks. The key ingredient in our architecture design is a new network block, which we name UniDual. It consists of a shared 2D spatial convolution followed by two parallel point-wise convolutional layers, one devoted to images and the other one used for videos. For video input, the point-wise filtering implements a temporal convolution. For image input, it performs a pixel-wise nonlinear transformation. Repeated stacking of such blocks gives rise to a network where images and videos undergo partially distinct execution pathways, unified by spatial convolutions (capturing commonalities in visual appearance) but separated by point-wise operations (modeling patterns specific to each modality). Extensive experiments on Kinetics and ImageNet demonstrate that our UniDual model jointly trained on these datasets yields substantial accuracy gains for both tasks, compared to 1) training separate models, 2) traditional multi-task learning and 3) the conventional framework of pretraining-followed-by-finetuning. On Kinetics, the UniDual architecture applied to a state-of-the-art video backbone model (R(2+1)D-152) yields an additional video@1 accuracy gain of 1.5%.
High accuracy video label prediction (classification) models are attributed to large scale data. These data could be frame feature sequences extracted by a pre-trained convolutional-neural-network, which promote the efficiency for creating models. Unsupervised solutions such as feature average pooling, as a simple label-independent parameter-free based method, has limited ability to represent the video. While the supervised methods, like RNN, can greatly improve the recognition accuracy. However, the video length is usually long, and there are hierarchical relationships between frames across events in the video, the performance of RNN based models are decreased. In this paper, we proposes a novel video classification method based on a deep convolutional graph neural network(DCGN). The proposed method utilize the characteristics of the hierarchical structure of the video, and performed multi-level feature extraction on the video frame sequence through the graph network, obtained a video representation re ecting the event semantics hierarchically. We test our model on YouTube-8M Large-Scale Video Understanding dataset, and the result outperforms RNN based benchmarks.