Due to the limited compute power and storage capabilities of edge platforms, ''streaming'' often provides a better VR experience compared to ''rendering''. Yet, achieving high-quality VR streaming faces two significant challenges, namely, bandwidth limitations and the need for real-time operation with high frames per second (FPS). Previous efforts have tended to prioritize either conserving bandwidth without real-time performance or ensuring real-time operation without substantial bandwidth savings. In this work, we incorporate the concept of ''stereo similarity'' to develop a novel real-time stereo video compression framework for streaming, called Pirate. Unlike the previously proposed approaches that rely on large machine learning-based models for synthesizing stereo pairs from both eyes with disparity maps (which can be impractical for most edge platforms due to their high computational cost), Pirate iteratively synthesizes the target eye view using only a single eye view and its corresponding disparity and optical flow information, with alternating left or right eye transmission. This enables us to generate target view at an extremely low computational cost, even under bandwidth constraints as low as 0.1 bits per pixel (bpp), while maintaining a high frame rate of 90 FPS. Our evaluations also reveal that, the proposed approach not only achieves real-time VR streaming with a 20%-40% reduction in bandwidth usage, but also maintains similar superior quality standards.

Repetitive action counting aims to count the number of repetitive actions in a video. The critical challenge of this task is to uncover the periodic pattern between repetitive actions by computing feature similarity between frames. However, existing methods only rely on the RGB feature of each frame to compute the feature similarity while neglecting the background change of repetitive actions. The abrupt background change may cause feature discrepancies of the same action moment and lead to errors in counting. To this end, we propose a two-branch framework, i.e., RGB and motion branches, with the motion branch complementing the RGB branch to enhance the foreground motion feature learning. Specifically, foreground motion features are highlighted with flow-guided attention on frame features. In addition, to alleviate the noise from moving background distractors and reinforce the periodic pattern, we propose a temporal self-similarity matrix reconstruction loss to improve the temporal correspondence between the same motion feature from different frames. Lastly, to make the motion feature effectively supplement the RGB feature, we present a novel variance-prompted loss weights generation technique to automatically generate dynamic loss weights for two branches in collaborative training. Extensive experiments are conducted on the RepCount and UCFRep datasets to verify our proposed method with state-of-the-art performance. Our method also achieves the best performance on the cross-dataset generalization experiment.

Video question answering aims to provide correct answers given complex videos and related questions, posting high requirements of the comprehension ability in both video and language processing. Existing works phrase this task as a multi-modal fusion process by aligning the video context with the whole question, ignoring the rich semantic details of nouns and verbs separately in the multi-modal reasoning process to derive the final answer. To fill this gap, in addition to the semantic alignment of the whole sentence, we propose to disentangle the semantic understanding of language, and reason over the corresponding frame-level and motion-level video features. We design an unified multi-granularity language module of residual structure to adapt the semantic understanding at different granularity with context exchange, e.g., word-level and sentence-level. To enhance the holistic question understanding for answer prediction, we also design a contrastive sampling approach by selecting irrelevant questions as negative samples to break the intrinsic correlations between questions and answers within the dataset. Notably, our model is competent for both multiple-choice and open-ended video question answering. We further employ a pre-trained language model to retrieve relevant knowledge as candidate answer context to facilitate open-ended VideoQA. Extensive quantitative and qualitative experiments on four public datasets (NextQA, MSVD, MSRVTT, and TGIF-QA-R) demonstrate the effective and superior performance of our proposed model. Our code will be released upon the paper’s acceptance.

The long-tailed video recognition problem is especially challenging, as videos tend to be long and untrimmed, and each video may contain multiple classes, causing frame-level class imbalance. The previous method tackles the long-tailed video recognition only through frame-level sampling for class re-balance without distinguishing the frame-level feature representation between head and tail classes. To improve the frame-level feature representation of tail classes, we modulate the frame-level features with an auxiliary distillation loss to reduce the distribution distance between head and tail classes. Moreover, we design a mixture-of-experts framework with two different expert designs, i.e., the first expert with an attention-based classification network handling the original long-tailed distribution, and the second expert dealing with the re-balanced distribution from class-balanced sampling. Notably, in the second expert, we specifically focus on the frames unsolved by the first expert through designing a complementary frame selection module, which inherits the attention weights from the first expert and selects frames with low attention weights, and we also enhance the motion feature representation for these selected frames. To highlight the multi-label challenge in long-tailed video recognition, we create two additional benchmarks based on Charades and CharadesEgo videos with the multi-label property, called CharadesLT and CharadesEgoLT. Extensive experiments are conducted on the existing long-tailed video benchmark VideoLT and the two new benchmarks to verify the effectiveness of our proposed method with state-of-the-art performance. The code and proposed benchmarks are released at https://github.com/VisionLanguageLab/MEID.

Temporal sentence grounding is a challenging task that aims to localize the semantic corresponding segment from the untrimmed video according to the given query language description. Existing methods either utilize a cross-modal matching architecture following a scan-and-rank pipeline or directly predict the probabilities of being the target boundary for each frame based on the entire video content. However, such methods are weak when some of the critical semantic concepts in the query are actually relevant to multiple video segments or the desired video segment contains a query-irrelevant scene due to ignoring query semantic concepts and local and global cross-modal context. In this paper, we propose a novel semantic-aware graph calibration network (SaGCN) to address the issues mentioned above. Specifically, we first introduce a semantic-aware local relational graph module to capture the inherent relationships among the specific semantic concept relevant local contextual information for fine-grained cross-modal information interactions. Then, a semantic-aware global relational graph module is derived for global contextual information integration and achieving cross-modal alignment. Finally, an attention-based calibration module is designed for eliminating the irrelevant information maintained in the visual modality under the guidance of query description. Extensive experiments verify the effectiveness of our proposed SaGCN on two widely used datasets (Charades-STA and TACoS), in which we achieve significant and consistent improvement compared to the state-of-the-art approaches.

Zhekun Luo, Shalini Ghosh, Devin Guillory, Keizo Kato, Trevor Darrell, Huijuan Xu. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2022.

Meta-learning has been the most common framework for few-shot learning in recent years. It learns the model from collections of few-shot classification tasks, which is believed to have a key advantage of making the training objective consistent with the testing objective. However, some recent works report that by training for whole-classification, i.e. classification on the whole label-set, it can get comparable or even better embedding than many meta-learning algorithms. The edge between these two lines of works has yet been underexplored, and the effectiveness of meta-learning in few-shot learning remains unclear. In this paper, we explore a simple process: meta-learning over a whole-classification pre-trained model on its evaluation metric. We observe this simple method achieves competitive performance to state-of-the-art methods on standard bench-marks. Our further analysis shed some light on understanding the trade-offs between the meta-learning objective and the whole-classification objective in few-shot learning. Our code is available at https://github.com/yinboc/few-shot-meta-baseline.

Training temporal action detection in videos requires large amounts of labeled data, yet such annotation is expensive to collect. Incorporating unlabeled or weakly-labeled data to train action detection model could help reduce annotation cost. In this work, we first introduce the Semi-supervised Action Detection (SSAD) task with a mixture of labeled and unlabeled data and analyze different types of errors in the proposed SSAD baselines which are directly adapted from the semi-supervised classification literature. Identifying that the main source of error is action incompleteness (i.e., missing parts of actions), we alleviate it by designing an unsupervised foreground attention (UFA) module utilizing the conditional independence between foreground and background motion. Then we incorporate weakly-labeled data into SSAD and propose Omni-supervised Action Detection (OSAD) with three levels of supervision. To overcome the accompanying action-context confusion problem in OSAD baselines, an information bottleneck (IB) is designed to suppress the scene information in non-action frames while preserving the action information. We extensively benchmark against the baselines for SSAD and OSAD on our created data splits in THUMOS14 and ActivityNet1.2, and demonstrate the effectiveness of the proposed UFA and IB methods. Lastly, the benefit of our full OSAD-IB model under limited annotation budgets is shown by exploring the optimal annotation strategy for labeled, unlabeled and weakly-labeled data.1

The goal of weakly-supervised video moment retrieval is to localize the video segment most relevant to a description without access to temporal annotations during training. Prior work uses co-attention mechanisms to understand relationships between the vision and language data, but they lack contextual information between video frames that can be useful to determine how well a segment relates to the query. To address this, we propose an efficient Latent Graph Co-Attention Network (LoGAN) that exploits fine-grained frame-by-word interactions to jointly reason about the correspondences between all possible pairs of frames, providing context cues absent in prior work. Experiments on the DiDeMo and Charades-STA datasets demonstrate the effectiveness of our approach, where we improve Recall@1 by 520% over prior weakly-supervised methods, even boasting an 11% gain over strongly-supervised methods on DiDeMo, while also using significantly fewer model parameters than other co-attention mechanisms.

Human action is naturally compositional: humans can easily recognize and perform actions with objects that are different from those used in training demonstrations. In this paper, we study the compositionality of action by looking into the dynamics of subject-object interactions. We propose a novel model which can explicitly reason about the geometric relations between constituent objects and an agent performing an action. To train our model, we collect dense object box annotations on the Something-Something dataset. We propose a novel compositional action recognition task where the training combinations of verbs and nouns do not overlap with the test set. The novel aspects of our model are applicable to activities with prominent object interaction dynamics and to objects which can be tracked using state-of-the-art approaches; for activities without clearly defined spatial object-agent interactions, we rely on baseline scene-level spatio-temporal representations. We show the effectiveness of our approach not only on the proposed compositional action recognition task but also in a few-shot compositional setting which requires the model to generalize across both object appearance and action category.

Supervised learning requires a large amount of training data, limiting its application where labeled data is scarce. To compensate for data scarcity, one possible method is to utilize auxiliary tasks to provide additional supervision for the main task. Assigning and optimizing the importance weights for different auxiliary tasks remains an crucial and largely understudied research question. In this work, we propose a method to automatically reweight auxiliary tasks in order to reduce the data requirement on the main task. Specifically, we formulate the weighted likelihood function of auxiliary tasks as a surrogate prior for the main task. By adjusting the auxiliary task weights to minimize the divergence between the surrogate prior and the true prior of the main task, we obtain a more accurate prior estimation, achieving the goal of minimizing the required amount of training data for the main task and avoiding a costly grid search. In multiple experimental settings (e.g. semi-supervised learning, multi-label classification), we demonstrate that our algorithm can effectively utilize limited labeled data of the main task with the benefit of auxiliary tasks compared with previous task reweighting methods. We also show that under extreme cases with only a few extra examples (e.g. few-shot domain adaptation), our algorithm results in significant improvement over the baseline. Our code and video is available at https://sites.google.com/view/auxiliary-task-reweighting.

Video generation is an inherently challenging task, as it requires modeling realistic temporal dynamics as well as spatial content. Existing methods entangle the two intrinsically different tasks of motion and content creation in a single generator network, but this approach struggles to simultaneously generate plausible motion and content. To improve motion modeling in video generation task, we propose a two-stream model that disentangles motion generation from content generation, called a Two-Stream Variational Adversarial Network (TwoStreamVAN). Given an action label and a noise vector, our model is able to create clear and consistent motion, and thus yields photorealistic videos. The key idea is to progressively generate and fuse multi-scale motion with its corresponding spatial content. Our model significantly outperforms existing methods on the standard Weizmann Human Action, MUG Facial Expression and VoxCeleb datasets, as well as our new dataset of diverse human actions with challenging and complex motion. Our code is available at https://github.com/sunxm2357/TwoStreamVAN/.

Pedestrian detection in the wild remains a challenging problem especially when the scene contains significant occlusion and/or low resolution of the pedestrians to be detected. Existing methods are unable to adapt to these difficult cases while maintaining acceptable performance. In this paper we propose a novel feature learning model, referred to as CircleNet, to achieve feature adaptation by mimicking the process humans looking at low resolution and occluded objects: focusing on it again, at a finer scale, if the object can not be identified clearly for the first time. CircleNet is implemented as a set of feature pyramids and uses weight sharing path augmentation for better feature fusion. It targets at reciprocating feature adaptation and iterative object detection using multiple top-down and bottom-up pathways. To take full advantage of the feature adaptation capability in CircleNet, we design an instance decomposition training strategy to focus on detecting pedestrian instances of various resolutions and different occlusion levels in each cycle. Specifically, CircleNet implements feature ensemble with the idea of hard negative boosting in an end-to-end manner. Experiments on two pedestrian detection datasets, Caltech and CityPersons, show that CircleNet improves the performance of occluded and low-resolution pedestrians with significant margins while maintaining good performance on normal instances.

We address the problem of text-based activity retrieval in video. Given a sentence describing an activity, our task is to retrieve matching clips from an untrimmed video. To capture the inherent structures present in both text and video, we introduce a multilevel model that integrates vision and language features earlier and more tightly than prior work. First, we inject text features early on when generating clip proposals, to help eliminate unlikely clips and thus speed up processing and boost performance. Second, to learn a fine-grained similarity metric for retrieval, we use visual features to modulate the processing of query sentences at the word level in a recurrent neural network. A multi-task loss is also employed by adding query re-generation as an auxiliary task. Our approach significantly outperforms prior work on two challenging benchmarks: Charades-STA and ActivityNet Captions.

Discriminative region responses residing inside an object instance can be extracted from networks trained with image-level label supervision. However, learning the full extent of pixel-level instance response in a weakly supervised manner remains unexplored. In this work, we tackle this challenging problem by using a novel instance extent filling approach. We first design a process to selectively collect pseudo supervision from noisy segment proposals obtained with previously published techniques. The pseudo supervision is used to learn a differentiable filling module that predicts a class-agnostic activation map for each instance given the image and an incomplete region response. We refer to the above maps as Instance Activation Maps (IAMs), which provide a fine-grained instance-level representation and allow instance masks to be extracted by lightweight CRF. Extensive experiments on the PASCAL VOC12 dataset show that our approach beats the state-of-the-art weakly supervised instance segmentation methods by a significant margin and increases the inference speed by an order of magnitude. Our method also generalizes well across domains and to unseen object categories. Without fine-tuning for the specific tasks, our model trained on VOC12 dataset (20 classes) obtains top performance for weakly supervised object localization on the CUB dataset (200 classes) and achieves competitive results on three widely used salient object detection benchmarks.

Dense video captioning is a fine-grained video understanding task that involves two sub-problems: localizing distinct events in a long video stream, and generating captions for the localized events. We propose the Joint Event Detection and Description Network (JEDDi-Net), which solves the dense video captioning task in an end-to-end fashion. Our model continuously encodes the input video stream with three-dimensional convolutional layers, proposes variable-length temporal events based on pooled features, and generates their captions. Proposal features are extracted within each proposal segment through 3D Segment-of-Interest pooling from shared video feature encoding. In order to explicitly model temporal relationships between visual events and their captions in a single video, we also propose a two-level hierarchical captioning module that keeps track of context. On the large-scale ActivityNet Captions dataset, JEDDi-Net demonstrates improved results as measured by standard metrics. We also present the first dense captioning results on the TACoS-MultiLevel dataset.

We address the problem of temporal activity detection in continuous, untrimmed video streams. This is a difficult task that requires extracting meaningful spatio-temporal features to capture activities, accurately localizing the start and end times of each activity. We introduce a new model, Region Convolutional 3D Network (R-C3D), which encodes the video streams using a three-dimensional fully convolutional network, then generates candidate temporal regions containing activities and finally classifies selected regions into specific activities. Computation is saved due to the sharing of convolutional features between the proposal and the classification pipelines. We further improve the detection performance by efficiently integrating an optical flow based motion stream with the original RGB stream. The two-stream network is jointly optimized by fusing the flow and RGB feature maps at different levels. Additionally, the training stage incorporates an online hard example mining strategy to address the extreme foreground-background imbalance typically observed in any detection pipeline. Instead of heuristically sampling the candidate segments for the final activity classification stage, we rank them according to their performance and only select the worst performers to update the model. This improves the model without heavy hyper-parameter tuning. Extensive experiments on three benchmark datasets are carried out to show superior performance over existing temporal activity detection methods. Our model achieves state-of-the-art results on the THUMOS'14 and Charades datasets. We further demonstrate that our model is a general temporal activity detection framework that does not rely on assumptions about particular dataset properties by evaluating our approach on the ActivityNet dataset.

Video prediction is a challenging problem due to the highly complex variation of video appearance and motions. Traditional methods that directly predict pixel values often result in blurring and artifacts. Furthermore, cumulative errors can lead to a sharp drop of prediction quality in long-term prediction. To alleviate the above problems, we propose a novel edge guided video prediction network, which firstly models the dynamic of frame edges and predicts the future frame edges, then generates the future frames under the guidance of the obtained future frame edges. Specifically, our network consists of two modules that are ConvLSTM based edge prediction module and the edge guided frames generation module. The whole network is differentiable and can be trained end-to-end without any supervision effort. Extensive experiments on KTH human action dataset and challenging autonomous driving KITTI dataset demonstrate that our method achieves better results than state-of-the-art methods especially in long-term video predictions.

We address the problem of activity detection in continuous, untrimmed video streams. This is a difficult task that requires extracting meaningful spatio-temporal features to capture activities, accurately localizing the start and end times of each activity. We introduce a new model, Region Convolutional 3D Network (R-C3D), which encodes the video streams using a three-dimensional fully convolutional network, then generates candidate temporal regions containing activities, and finally classifies selected regions into specific activities. Computation is saved due to the sharing of convolutional features between the proposal and the classification pipelines. The entire model is trained end-to-end with jointly optimized localization and classification losses. R-C3D is faster than existing methods (569 frames per second on a single Titan X Maxwell GPU) and achieves state-of-the-art results on THUMOS'14. We further demonstrate that our model is a general activity detection framework that does not rely on assumptions about particular dataset properties by evaluating our approach on ActivityNet and Charades. Our code is available at http://ai.bu.edu/r-c3d/.

Subhashini Venugopalan, Huijuan Xu, Jeff Donahue, Marcus Rohrbach, Raymond Mooney, Kate Saenko. Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2015.