Effectively tackling the problem of temporal action localization (TAL) necessitates a visual representation that jointly pursues two confounding goals, i.e., fine-grained discrimination for temporal localization and sufficient visual invariance for action classification. We address this challenge by enriching both the local and global contexts in the popular two-stage temporal localization framework, where action proposals are first generated followed by action classification and temporal boundary regression. Our proposed model, dubbed ContextLoc, can be divided into three sub-networks: L-Net, G-Net and P-Net. L-Net enriches the local context via fine-grained modeling of snippet-level features, which is formulated as a query-and-retrieval process. G-Net enriches the global context via higher-level modeling of the video-level representation. In addition, we introduce a novel context adaptation module to adapt the global context to different proposals. P-Net further models the context-aware inter-proposal relations. We explore two existing models to be the P-Net in our experiments. The efficacy of our proposed method is validated by experimental results on the THUMOS14 (54.3\% at [email omitted]) and ActivityNet v1.3 (56.01\% at [email omitted]) datasets, which outperforms recent states of the art. Code is available at https://github.com/buxiangzhiren/ContextLoc.
Action recognition based on skeleton data has recently witnessed increasing attention and progress. State-of-the-art approaches adopting Graph Convolutional networks (GCNs) can effectively extract features on human skeletons relying on the pre-defined human topology. Despite associated progress, GCN-based methods have difficulties to generalize across domains, especially with different human topological structures. In this context, we introduce UNIK, a novel skeleton-based action recognition method that is not only effective to learn spatio-temporal features on human skeleton sequences but also able to generalize across datasets. This is achieved by learning an optimal dependency matrix from the uniform distribution based on a multi-head attention mechanism. Subsequently, to study the cross-domain generalizability of skeleton-based action recognition in real-world videos, we re-evaluate state-of-the-art approaches as well as the proposed UNIK in light of a novel Posetics dataset. This dataset is created from Kinetics-400 videos by estimating, refining and filtering poses. We provide an analysis on how much performance improves on smaller benchmark datasets after pre-training on Posetics for the action classification task. Experimental results show that the proposed UNIK, with pre-training on Posetics, generalizes well and outperforms state-of-the-art when transferred onto four target action classification datasets: Toyota Smarthome, Penn Action, NTU-RGB+D 60 and NTU-RGB+D 120.
Recognizing Activities of Daily Living (ADL) is a vital process for intelligent assistive robots, but collecting large annotated datasets requires time-consuming temporal labeling and raises privacy concerns, e.g., if the data is collected in a real household. In this work, we explore the concept of constructing training examples for ADL recognition by playing life simulation video games and introduce the SIMS4ACTION dataset created with the popular commercial game THE SIMS 4. We build Sims4Action by specifically executing actions-of-interest in a "top-down" manner, while the gaming circumstances allow us to freely switch between environments, camera angles and subject appearances. While ADL recognition on gaming data is interesting from the theoretical perspective, the key challenge arises from transferring it to the real-world applications, such as smart-homes or assistive robotics. To meet this requirement, Sims4Action is accompanied with a GamingToReal benchmark, where the models are evaluated on real videos derived from an existing ADL dataset. We integrate two modern algorithms for video-based activity recognition in our framework, revealing the value of life simulation video games as an inexpensive and far less intrusive source of training data. However, our results also indicate that tasks involving a mixture of gaming and real data are challenging, opening a new research direction. We will make our dataset publicly available at https://github.com/aroitberg/sims4action.
Humans perceive the world by concurrently processing and fusing high-dimensional inputs from multiple modalities such as vision and audio. Machine perception models, in stark contrast, are typically modality-specific and optimised for unimodal benchmarks, and hence late-stage fusion of final representations or predictions from each modality (`late-fusion') is still a dominant paradigm for multimodal video classification. Instead, we introduce a novel transformer based architecture that uses `fusion bottlenecks' for modality fusion at multiple layers. Compared to traditional pairwise self-attention, our model forces information between different modalities to pass through a small number of bottleneck latents, requiring the model to collate and condense the most relevant information in each modality and only share what is necessary. We find that such a strategy improves fusion performance, at the same time reducing computational cost. We conduct thorough ablation studies, and achieve state-of-the-art results on multiple audio-visual classification benchmarks including Audioset, Epic-Kitchens and VGGSound. All code and models will be released.
The vision community is witnessing a modeling shift from CNNs to Transformers, where pure Transformer architectures have attained top accuracy on the major video recognition benchmarks. These video models are all built on Transformer layers that globally connect patches across the spatial and temporal dimensions. In this paper, we instead advocate an inductive bias of locality in video Transformers, which leads to a better speed-accuracy trade-off compared to previous approaches which compute self-attention globally even with spatial-temporal factorization. The locality of the proposed video architecture is realized by adapting the Swin Transformer designed for the image domain, while continuing to leverage the power of pre-trained image models. Our approach achieves state-of-the-art accuracy on a broad range of video recognition benchmarks, including on action recognition (84.9 top-1 accuracy on Kinetics-400 and 86.1 top-1 accuracy on Kinetics-600 with ~20x less pre-training data and ~3x smaller model size) and temporal modeling (69.6 top-1 accuracy on Something-Something v2). The code and models will be made publicly available at https://github.com/SwinTransformer/Video-Swin-Transformer.
In this paper, we introduce a novel visual representation learning which relies on a handful of adaptively learned tokens, and which is applicable to both image and video understanding tasks. Instead of relying on hand-designed splitting strategies to obtain visual tokens and processing a large number of densely sampled patches for attention, our approach learns to mine important tokens in visual data. This results in efficiently and effectively finding a few important visual tokens and enables modeling of pairwise attention between such tokens, over a longer temporal horizon for videos, or the spatial content in images. Our experiments demonstrate strong performance on several challenging benchmarks for both image and video recognition tasks. Importantly, due to our tokens being adaptive, we accomplish competitive results at significantly reduced compute amount. We obtain comparable results to the state-of-the-arts on ImageNet while being computationally more efficient. We also confirm the effectiveness of the approach on multiple video datasets, including Kinetics-400, Kinetics-600, Charades, and AViD. The code is available at: https://github.com/google-research/scenic/tree/main/scenic/projects/token_learner
Video understanding relies on perceiving the global content and modeling its internal connections (e.g., causality, movement, and spatio-temporal correspondence). To learn these interactions, we apply a mask-then-predict pre-training task on discretized video tokens generated via VQ-VAE. Unlike language, where the text tokens are more independent, neighboring video tokens typically have strong correlations (e.g., consecutive video frames usually look very similar), and hence uniformly masking individual tokens will make the task too trivial to learn useful representations. To deal with this issue, we propose a block-wise masking strategy where we mask neighboring video tokens in both spatial and temporal domains. We also add an augmentation-free contrastive learning method to further capture the global content by predicting whether the video clips are sampled from the same video. We pre-train our model on uncurated videos and show that our pre-trained model can reach state-of-the-art results on several video understanding datasets (e.g., SSV2, Diving48). Lastly, we provide detailed analyses on model scalability and pre-training method design. Code is released at https://github.com/airsplay/vimpac.
Understanding the semantics of human movement -- the what, how and why of the movement -- is an important problem that requires datasets of human actions with semantic labels. Existing datasets take one of two approaches. Large-scale video datasets contain many action labels but do not contain ground-truth 3D human motion. Alternatively, motion-capture (mocap) datasets have precise body motions but are limited to a small number of actions. To address this, we present BABEL, a large dataset with language labels describing the actions being performed in mocap sequences. BABEL consists of action labels for about 43 hours of mocap sequences from AMASS. Action labels are at two levels of abstraction -- sequence labels describe the overall action in the sequence, and frame labels describe all actions in every frame of the sequence. Each frame label is precisely aligned with the duration of the corresponding action in the mocap sequence, and multiple actions can overlap. There are over 28k sequence labels, and 63k frame labels in BABEL, which belong to over 250 unique action categories. Labels from BABEL can be leveraged for tasks like action recognition, temporal action localization, motion synthesis, etc. To demonstrate the value of BABEL as a benchmark, we evaluate the performance of models on 3D action recognition. We demonstrate that BABEL poses interesting learning challenges that are applicable to real-world scenarios, and can serve as a useful benchmark of progress in 3D action recognition. The dataset, baseline method, and evaluation code is made available, and supported for academic research purposes at https://babel.is.tue.mpg.de/.
In video transformers, the time dimension is often treated in the same way as the two spatial dimensions. However, in a scene where objects or the camera may move, a physical point imaged at one location in frame $t$ may be entirely unrelated to what is found at that location in frame $t+k$. These temporal correspondences should be modeled to facilitate learning about dynamic scenes. To this end, we propose a new drop-in block for video transformers -- trajectory attention -- that aggregates information along implicitly determined motion paths. We additionally propose a new method to address the quadratic dependence of computation and memory on the input size, which is particularly important for high resolution or long videos. While these ideas are useful in a range of settings, we apply them to the specific task of video action recognition with a transformer model and obtain state-of-the-art results on the Kinetics, Something--Something V2, and Epic-Kitchens datasets. Code and models are available at: https://github.com/facebookresearch/Motionformer
This paper is on video recognition using Transformers. Very recent attempts in this area have demonstrated promising results in terms of recognition accuracy, yet they have been also shown to induce, in many cases, significant computational overheads due to the additional modelling of the temporal information. In this work, we propose a Video Transformer model the complexity of which scales linearly with the number of frames in the video sequence and hence induces no overhead compared to an image-based Transformer model. To achieve this, our model makes two approximations to the full space-time attention used in Video Transformers: (a) It restricts time attention to a local temporal window and capitalizes on the Transformer's depth to obtain full temporal coverage of the video sequence. (b) It uses efficient space-time mixing to attend jointly spatial and temporal locations without inducing any additional cost on top of a spatial-only attention model. We also show how to integrate 2 very lightweight mechanisms for global temporal-only attention which provide additional accuracy improvements at minimal computational cost. We demonstrate that our model produces very high recognition accuracy on the most popular video recognition datasets while at the same time being significantly more efficient than other Video Transformer models. Code will be made available.
3D convolution is powerful for video classification but often computationally expensive, recent studies mainly focus on decomposing it on spatial-temporal and/or channel dimensions. Unfortunately, most approaches fail to achieve a preferable balance between convolutional efficiency and feature-interaction sufficiency. For this reason, we propose a concise and novel Channel Tensorization Network (CT-Net), by treating the channel dimension of input feature as a multiplication of K sub-dimensions. On one hand, it naturally factorizes convolution in a multiple dimension way, leading to a light computation burden. On the other hand, it can effectively enhance feature interaction from different channels, and progressively enlarge the 3D receptive field of such interaction to boost classification accuracy. Furthermore, we equip our CT-Module with a Tensor Excitation (TE) mechanism. It can learn to exploit spatial, temporal and channel attention in a high-dimensional manner, to improve the cooperative power of all the feature dimensions in our CT-Module. Finally, we flexibly adapt ResNet as our CT-Net. Extensive experiments are conducted on several challenging video benchmarks, e.g., Kinetics-400, Something-Something V1 and V2. Our CT-Net outperforms a number of recent SOTA approaches, in terms of accuracy and/or efficiency. The codes and models will be available on https://github.com/Andy1621/CT-Net.
We introduce Continual 3D Convolutional Neural Networks (Co3D CNNs), a new computational formulation of spatio-temporal 3D CNNs, in which videos are processed frame-by-frame rather than by clip. In online tasks demanding frame-wise predictions, Co3D CNNs dispense with the computational redundancies of regular 3D CNNs, namely the repeated convolutions over frames, which appear in overlapping clips. We show that Continual 3D CNNs can reuse preexisting 3D-CNN weights to reduce the per-prediction floating point operations (FLOPs) in proportion to the temporal receptive field while retaining similar memory requirements and accuracy. This is validated with multiple models on Kinetics-400 and Charades with remarkable results: CoX3D models attain state-of-the-art complexity/accuracy trade-offs on Kinetics-400 with 12.1-15.3x reductions of FLOPs and 2.3-3.8% improvements in accuracy compared to regular X3D models while reducing peak memory consumption by up to 48%. Moreover, we investigate the transient response of Co3D CNNs at start-up and perform extensive benchmarks of on-hardware processing characteristics for publicly available 3D CNNs.
Many attempts have been made towards combining RGB and 3D poses for the recognition of Activities of Daily Living (ADL). ADL may look very similar and often necessitate to model fine-grained details to distinguish them. Because the recent 3D ConvNets are too rigid to capture the subtle visual patterns across an action, this research direction is dominated by methods combining RGB and 3D Poses. But the cost of computing 3D poses from RGB stream is high in the absence of appropriate sensors. This limits the usage of aforementioned approaches in real-world applications requiring low latency. Then, how to best take advantage of 3D Poses for recognizing ADL? To this end, we propose an extension of a pose driven attention mechanism: Video-Pose Network (VPN), exploring two distinct directions. One is to transfer the Pose knowledge into RGB through a feature-level distillation and the other towards mimicking pose driven attention through an attention-level distillation. Finally, these two approaches are integrated into a single model, we call VPN++. We show that VPN++ is not only effective but also provides a high speed up and high resilience to noisy Poses. VPN++, with or without 3D Poses, outperforms the representative baselines on 4 public datasets. Code is available at https://github.com/srijandas07/vpnplusplus.
We introduce a weakly supervised method for representation learning based on aligning temporal sequences (e.g., videos) of the same process (e.g., human action). The main idea is to use the global temporal ordering of latent correspondences across sequence pairs as a supervisory signal. In particular, we propose a loss based on scoring the optimal sequence alignment to train an embedding network. Our loss is based on a novel probabilistic path finding view of dynamic time warping (DTW) that contains the following three key features: (i) the local path routing decisions are contrastive and differentiable, (ii) pairwise distances are cast as probabilities that are contrastive as well, and (iii) our formulation naturally admits a global cycle consistency loss that verifies correspondences. For evaluation, we consider the tasks of fine-grained action classification, few shot learning, and video synchronization. We report significant performance increases over previous methods. In addition, we report two applications of our temporal alignment framework, namely 3D pose reconstruction and fine-grained audio/visual retrieval.
Inspired by the fact that human eyes continue to develop tracking ability in early and middle childhood, we propose to use tracking as a proxy task for a computer vision system to learn the visual representations. Modelled on the Catch game played by the children, we design a Catch-the-Patch (CtP) game for a 3D-CNN model to learn visual representations that would help with video-related tasks. In the proposed pretraining framework, we cut an image patch from a given video and let it scale and move according to a pre-set trajectory. The proxy task is to estimate the position and size of the image patch in a sequence of video frames, given only the target bounding box in the first frame. We discover that using multiple image patches simultaneously brings clear benefits. We further increase the difficulty of the game by randomly making patches invisible. Extensive experiments on mainstream benchmarks demonstrate the superior performance of CtP against other video pretraining methods. In addition, CtP-pretrained features are less sensitive to domain gaps than those trained by a supervised action recognition task. When both trained on Kinetics-400, we are pleasantly surprised to find that CtP-pretrained representation achieves much higher action classification accuracy than its fully supervised counterpart on Something-Something dataset. Code is available online: github.com/microsoft/CtP.
We present Multiscale Vision Transformers (MViT) for video and image recognition, by connecting the seminal idea of multiscale feature hierarchies with transformer models. Multiscale Transformers have several channel-resolution scale stages. Starting from the input resolution and a small channel dimension, the stages hierarchically expand the channel capacity while reducing the spatial resolution. This creates a multiscale pyramid of features with early layers operating at high spatial resolution to model simple low-level visual information, and deeper layers at spatially coarse, but complex, high-dimensional features. We evaluate this fundamental architectural prior for modeling the dense nature of visual signals for a variety of video recognition tasks where it outperforms concurrent vision transformers that rely on large scale external pre-training and are 5-10x more costly in computation and parameters. We further remove the temporal dimension and apply our model for image classification where it outperforms prior work on vision transformers. Code is available at: https://github.com/facebookresearch/SlowFast
We present a framework for learning multimodal representations from unlabeled data using convolution-free Transformer architectures. Specifically, our Video-Audio-Text Transformer (VATT) takes raw signals as inputs and extracts multimodal representations that are rich enough to benefit a variety of downstream tasks. We train VATT end-to-end from scratch using multimodal contrastive losses and evaluate its performance by the downstream tasks of video action recognition, audio event classification, image classification, and text-to-video retrieval. Furthermore, we study a modality-agnostic, single-backbone Transformer by sharing weights among the three modalities. We show that the convolution-free VATT outperforms state-of-the-art ConvNet-based architectures in the downstream tasks. Especially, VATT's vision Transformer achieves the top-1 accuracy of 82.1% on Kinetics-400, 83.6% on Kinetics-600, 72.7% on Kinetics-700, and 41.1% on Moments in Time, new records while avoiding supervised pre-training. Transferring to image classification leads to 78.7% top-1 accuracy on ImageNet compared to 64.7% by training the same Transformer from scratch, showing the generalizability of our model despite the domain gap between videos and images. VATT's audio Transformer also sets a new record on waveform-based audio event recognition by achieving the mAP of 39.4% on AudioSet without any supervised pre-training. VATT's source code is publicly available.
This work strives for the classification and localization of human actions in videos, without the need for any labeled video training examples. Where existing work relies on transferring global attribute or object information from seen to unseen action videos, we seek to classify and spatio-temporally localize unseen actions in videos from image-based object information only. We propose three spatial object priors, which encode local person and object detectors along with their spatial relations. On top we introduce three semantic object priors, which extend semantic matching through word embeddings with three simple functions that tackle semantic ambiguity, object discrimination, and object naming. A video embedding combines the spatial and semantic object priors. It enables us to introduce a new video retrieval task that retrieves action tubes in video collections based on user-specified objects, spatial relations, and object size. Experimental evaluation on five action datasets shows the importance of spatial and semantic object priors for unseen actions. We find that persons and objects have preferred spatial relations that benefit unseen action localization, while using multiple languages and simple object filtering directly improves semantic matching, leading to state-of-the-art results for both unseen action classification and localization.
In video data, busy motion details from moving regions are conveyed within a specific frequency bandwidth in the frequency domain. Meanwhile, the rest of the frequencies of video data are encoded with quiet information with substantial redundancy, which causes low processing efficiency in existing video models that take as input raw RGB frames. In this paper, we consider allocating intenser computation for the processing of the important busy information and less computation for that of the quiet information. We design a trainable Motion Band-Pass Module (MBPM) for separating busy information from quiet information in raw video data. By embedding the MBPM into a two-pathway CNN architecture, we define a Busy-Quiet Net (BQN). The efficiency of BQN is determined by avoiding redundancy in the feature space processed by the two pathways: one operating on Quiet features of low-resolution, while the other processes Busy features. The proposed BQN outperforms many recent video processing models on Something-Something V1, Kinetics400, UCF101 and HMDB51 datasets.
We present pure-transformer based models for video classification, drawing upon the recent success of such models in image classification. Our model extracts spatio-temporal tokens from the input video, which are then encoded by a series of transformer layers. In order to handle the long sequences of tokens encountered in video, we propose several, efficient variants of our model which factorise the spatial- and temporal-dimensions of the input. Although transformer-based models are known to only be effective when large training datasets are available, we show how we can effectively regularise the model during training and leverage pretrained image models to be able to train on comparatively small datasets. We conduct thorough ablation studies, and achieve state-of-the-art results on multiple video classification benchmarks including Kinetics 400 and 600, Epic Kitchens, Something-Something v2 and Moments in Time, outperforming prior methods based on deep 3D convolutional networks. To facilitate further research, we release code at https://github.com/google-research/scenic/tree/main/scenic/projects/vivit