To understand the world, we humans constantly need to relate the present to the past, and put events in context. In this paper, we enable existing video models to do the same. We propose a long-term feature bank---supportive information extracted over the entire span of a video---to augment state-of-the-art video models that otherwise would only view short clips of 2-5 seconds. Our experiments demonstrate that augmenting 3D convolutional networks with a long-term feature bank yields state-of-the-art results on three challenging video datasets: AVA, EPIC-Kitchens, and Charades.
The explosive growth in video streaming gives rise to challenges on performing video understanding at high accuracy and low computation cost. Conventional 2D CNNs are computationally cheap but cannot capture temporal relationships; 3D CNN based methods can achieve good performance but are computationally intensive, making it expensive to deploy. In this paper, we propose a generic and effective Temporal Shift Module (TSM) that enjoys both high efficiency and high performance. Specifically, it can achieve the performance of 3D CNN but maintain 2D CNN's complexity. TSM shifts part of the channels along the temporal dimension; thus facilitate information exchanged among neighboring frames. It can be inserted into 2D CNNs to achieve temporal modeling at zero computation and zero parameters. We also extended TSM to online setting, which enables real-time low-latency online video recognition and video object detection. TSM is accurate and efficient: it ranks the first place on the Something-Something leaderboard upon publication; on Jetson Nano and Galaxy Note8, it achieves a low latency of 13ms and 35ms for online video recognition. The code is available at: https://github.com/mit-han-lab/temporal-shift-module.
In this paper, we propose a convolutional layer inspired by optical flow algorithms to learn motion representations. Our representation flow layer is a fully-differentiable layer designed to capture the `flow' of any representation channel within a convolutional neural network for action recognition. Its parameters for iterative flow optimization are learned in an end-to-end fashion together with the other CNN model parameters, maximizing the action recognition performance. Furthermore, we newly introduce the concept of learning `flow of flow' representations by stacking multiple representation flow layers. We conducted extensive experimental evaluations, confirming its advantages over previous recognition models using traditional optical flows in both computational speed and performance. Code/models available here: https://piergiaj.github.io/rep-flow-site/
We introduce modifications to state-of-the-art approaches to aggregating local video descriptors by using attention mechanisms and function approximations. Rather than using ensembles of existing architectures, we provide an insight on creating new architectures. We demonstrate our solutions in the "The 2nd YouTube-8M Video Understanding Challenge", by using frame-level video and audio descriptors. We obtain testing accuracy similar to the state of the art, while meeting budget constraints, and touch upon strategies to improve the state of the art. Model implementations are available in https://github.com/pomonam/LearnablePoolingMethods.
This paper presents our 7th place solution to the second YouTube-8M video understanding competition which challenges participates to build a constrained-size model to classify millions of YouTube videos into thousands of classes. Our final model consists of four single models aggregated into one tensorflow graph. For each single model, we use the same network architecture as in the winning solution of the first YouTube-8M video understanding competition, namely Gated NetVLAD. We train the single models separately in tensorflow's default float32 precision, then replace weights with float16 precision and ensemble them in the evaluation and inference stages., achieving 48.5% compression rate without loss of precision. Our best model achieved 88.324% GAP on private leaderboard. The code is publicly available at https://github.com/boliu61/youtube-8m
Despite the recent progress in video understanding and the continuous rate of improvement in temporal action localization throughout the years, it is still unclear how far (or close?) we are to solving the problem. To this end, we introduce a new diagnostic tool to analyze the performance of temporal action detectors in videos and compare different methods beyond a single scalar metric. We exemplify the use of our tool by analyzing the performance of the top rewarded entries in the latest ActivityNet action localization challenge. Our analysis shows that the most impactful areas to work on are: strategies to better handle temporal context around the instances, improving the robustness w.r.t. the instance absolute and relative size, and strategies to reduce the localization errors. Moreover, our experimental analysis finds the lack of agreement among annotator is not a major roadblock to attain progress in the field. Our diagnostic tool is publicly available to keep fueling the minds of other researchers with additional insights about their algorithms.
Recent two-stream deep Convolutional Neural Networks (ConvNets) have made significant progress in recognizing human actions in videos. Despite their success, methods extending the basic two-stream ConvNet have not systematically explored possible network architectures to further exploit spatiotemporal dynamics within video sequences. Further, such networks often use different baseline two-stream networks. Therefore, the differences and the distinguishing factors between various methods using Recurrent Neural Networks (RNN) or convolutional networks on temporally-constructed feature vectors (Temporal-ConvNet) are unclear. In this work, we first demonstrate a strong baseline two-stream ConvNet using ResNet-101. We use this baseline to thoroughly examine the use of both RNNs and Temporal-ConvNets for extracting spatiotemporal information. Building upon our experimental results, we then propose and investigate two different networks to further integrate spatiotemporal information: 1) temporal segment RNN and 2) Inception-style Temporal-ConvNet. We demonstrate that using both RNNs (using LSTMs) and Temporal-ConvNets on spatiotemporal feature matrices are able to exploit spatiotemporal dynamics to improve the overall performance. However, each of these methods require proper care to achieve state-of-the-art performance; for example, LSTMs require pre-segmented data or else they cannot fully exploit temporal information. Our analysis identifies specific limitations for each method that could form the basis of future work. Our experimental results on UCF101 and HMDB51 datasets achieve state-of-the-art performances, 94.1% and 69.0%, respectively, without requiring extensive temporal augmentation.
We present a general approach to video understanding, inspired by semantic transfer techniques that have been successfully used for 2D image analysis. Our method considers a video to be a 1D sequence of clips, each one associated with its own semantics. The nature of these semantics -- natural language captions or other labels -- depends on the task at hand. A test video is processed by forming correspondences between its clips and the clips of reference videos with known semantics, following which, reference semantics can be transferred to the test video. We describe two matching methods, both designed to ensure that (a) reference clips appear similar to test clips and (b), taken together, the semantics of the selected reference clips is consistent and maintains temporal coherence. We use our method for video captioning on the LSMDC'16 benchmark, video summarization on the SumMe and TVSum benchmarks, Temporal Action Detection on the Thumos2014 benchmark, and sound prediction on the Greatest Hits benchmark. Our method not only surpasses the state of the art, in four out of five benchmarks, but importantly, it is the only single method we know of that was successfully applied to such a diverse range of tasks.
Pixels operate locally. Superpixels have some potential to collect information across many pixels; supervoxels have more potential by implicitly operating across time. In this paper, we explore this well established notion thoroughly analyzing how supervoxels can be used in place of and in conjunction with other means of aggregating information across space-time. Focusing on the problem of strictly unsupervised video object segmentation, we devise a method called supervoxel gerrymandering that links masks of foregroundness and backgroundness via local and non-local consensus measures. We pose and answer a series of critical questions about the ability of supervoxels to adequately sway local voting; the questions regard type and scale of supervoxels as well as local versus non-local consensus, and the questions are posed in a general way so as to impact the broader knowledge of the use of supervoxels in video understanding. We work with the DAVIS dataset and find that our analysis yields an unsupervised method that outperforms all other known unsupervised methods and even many supervised ones.
This article describes the final solution of team monkeytyping, who finished in second place in the YouTube-8M video understanding challenge. The dataset used in this challenge is a large-scale benchmark for multi-label video classification. We extend the work in [1] and propose several improvements for frame sequence modeling. We propose a network structure called Chaining that can better capture the interactions between labels. Also, we report our approaches in dealing with multi-scale information and attention pooling. In addition, We find that using the output of model ensemble as a side target in training can boost single model performance. We report our experiments in bagging, boosting, cascade, and stacking, and propose a stacking algorithm called attention weighted stacking. Our final submission is an ensemble that consists of 74 sub models, all of which are listed in the appendix.
We present a solution to "Google Cloud and YouTube-8M Video Understanding Challenge" that ranked 5th place. The proposed model is an ensemble of three model families, two frame level and one video level. The training was performed on augmented dataset, with cross validation.
This paper introduces the system we developed for the Youtube-8M Video Understanding Challenge, in which a large-scale benchmark dataset was used for multi-label video classification. The proposed framework contains hierarchical deep architecture, including the frame-level sequence modeling part and the video-level classification part. In the frame-level sequence modelling part, we explore a set of methods including Pooling-LSTM (PLSTM), Hierarchical-LSTM (HLSTM), Random-LSTM (RLSTM) in order to address the problem of large amount of frames in a video. We also introduce two attention pooling methods, single attention pooling (ATT) and multiply attention pooling (Multi-ATT) so that we can pay more attention to the informative frames in a video and ignore the useless frames. In the video-level classification part, two methods are proposed to increase the classification performance, i.e. Hierarchical-Mixture-of-Experts (HMoE) and Classifier Chains (CC). Our final submission is an ensemble consisting of 18 sub-models. In terms of the official evaluation metric Global Average Precision (GAP) at 20, our best submission achieves 0.84346 on the public 50% of test dataset and 0.84333 on the private 50% of test data.
We took part in the YouTube-8M Video Understanding Challenge hosted on Kaggle, and achieved the 10th place within less than one month's time. In this paper, we present an extensive analysis and solution to the underlying machine-learning problem based on frame-level data, where major challenges are identified and corresponding preliminary methods are proposed. It's noteworthy that, with merely the proposed strategies and uniformly-averaging multi-crop ensemble was it sufficient for us to reach our ranking. We also report the methods we believe to be promising but didn't have enough time to train to convergence. We hope this paper could serve, to some extent, as a review and guideline of the YouTube-8M multi-label video classification benchmark, inspiring future attempts and research.
This paper describes our solution for the video recognition task of the Google Cloud and YouTube-8M Video Understanding Challenge that ranked the 3rd place. Because the challenge provides pre-extracted visual and audio features instead of the raw videos, we mainly investigate various temporal modeling approaches to aggregate the frame-level features for multi-label video recognition. Our system contains three major components: two-stream sequence model, fast-forward sequence model and temporal residual neural networks. Experiment results on the challenging Youtube-8M dataset demonstrate that our proposed temporal modeling approaches can significantly improve existing temporal modeling approaches in the large-scale video recognition tasks. To be noted, our fast-forward LSTM with a depth of 7 layers achieves 82.75% in term of GAP@20 on the Kaggle Public test set.
Current methods for video analysis often extract frame-level features using pre-trained convolutional neural networks (CNNs). Such features are then aggregated over time e.g., by simple temporal averaging or more sophisticated recurrent neural networks such as long short-term memory (LSTM) or gated recurrent units (GRU). In this work we revise existing video representations and study alternative methods for temporal aggregation. We first explore clustering-based aggregation layers and propose a two-stream architecture aggregating audio and visual features. We then introduce a learnable non-linear unit, named Context Gating, aiming to model interdependencies among network activations. Our experimental results show the advantage of both improvements for the task of video classification. In particular, we evaluate our method on the large-scale multi-modal Youtube-8M v2 dataset and outperform all other methods in the Youtube 8M Large-Scale Video Understanding challenge.
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.
The state of the art in video understanding suffers from two problems: (1) The major part of reasoning is performed locally in the video, therefore, it misses important relationships within actions that span several seconds. (2) While there are local methods with fast per-frame processing, the processing of the whole video is not efficient and hampers fast video retrieval or online classification of long-term activities. In this paper, we introduce a network architecture that takes long-term content into account and enables fast per-video processing at the same time. The architecture is based on merging long-term content already in the network rather than in a post-hoc fusion. Together with a sampling strategy, which exploits that neighboring frames are largely redundant, this yields high-quality action classification and video captioning at up to 230 videos per second, where each video can consist of a few hundred frames. The approach achieves competitive performance across all datasets while being 10x to 80x faster than state-of-the-art methods.
대부분의 기존 비디오 이상 탐지기(video anomaly detector)는 RGB 프레임에만 의존하므로, 이상 이벤트를 식별하는 데 중요한 급격하거나 일시적인 모션 단서를 포착하는 능력이 제한됩니다. 본 연구에서는 원리 기반 가중 메커니즘을 통해 상호 보완적인 RGB 및 합성 모션 정보를 통합하는 프레임워크인 Uncertainty Weighted Image Event Fusion(IEF-VAD)을 제안합니다. 이 방법은 합성 모션 단서의 높은 분산 및 꼬리가 두꺼운(heavy-tailed) 특성을 스튜던트 t(Student's t) 가능도(likelihood)로 모델링하고, 이미지 모달리티가 모션 정보를 압도하는 것을 방지하기 위해 라플라스 근사(Laplace approximation)를 사용하여 값 수준 역분산 가중치를 계산하며, 잔여 교차 모달 노이즈를 억제하기 위해 반복적 정제(iterative refinement)를 수행합니다. 이러한 정식화는 모달리티 지배(modality dominance) 현상을 자주 겪는 크로스 어텐션(cross-attention)이나 게이팅 기반 접근 방식에 비해 더 균형 잡히고 신뢰할 수 있는 융합 과정을 제공합니다. 이벤트 카메라나 프레임 수준의 주석 없이도 IEF-VAD는 여러 실제 이상 탐지 벤치마크(benchmark)에서 새로운 최고 성능(SOTA)을 달성하며 개별 모달리티에 열화가 적용되어도 안정성을 유지합니다. 이러한 결과는 상호 보완적인 모션 단서를 추출하고 통합하는 것이 다양한 환경 전반에서 강건한 비디오 이해를 위한 효과적인 방향임을 보여줍니다.
XOV-Action은 오픈-보캐뷸러리 행동 인식 모델이 미지의 비디오 도메인에서 일반화에 실패한다는 문제를 해결하기 위해, 다양한 텍스트 설명으로 행동 개념을 풍부하게 학습하는 Diversified Elaboration Representation Learning과 씬 편향을 제거하는 Scene-Aware Video-text Alignment를 제안한다. 이를 평가하기 위한 크로스-도메인 벤치마크 XOVABench도 함께 공개하며, 실험에서 클로즈드-셋 및 오픈-셋 행동 카테고리 모두에 걸쳐 도메인 일반화 성능이 향상됨을 입증했다.
OmniFall은 소규모 연출 데이터셋에만 의존하던 낙상 감지 모델의 한계를 극복하기 위해, 8개의 연출 데이터셋·합성 영상·실제 사고 영상을 통합한 15k개 비디오(80시간) 규모의 통합 벤치마크를 제안한다. 평가 결과, 제로샷 멀티모달 대규모 언어 모델(MLLM)과 파인튜닝 모델 모두 낙상 이벤트 감지에서 유사한 성능을 보였지만, 임상적으로 중요한 '쓰러진 상태(fallen state)' 감지에서는 합성 데이터로 파인튜닝한 모델이 제로샷 모델보다 월등히 우수했다.