In this work we explore how the architecture proposed in [8], which expresses the processing steps of the classical Fisher vector pipeline approaches, i.e. dimensionality reduction by principal component analysis (PCA) projection, Gaussian mixture model (GMM) and Fisher vector descriptor extraction as network layers, can be modified into a hybrid network that combines the benefits of both unsupervised and supervised training methods, resulting in a model that learns a semi-supervised Fisher vector descriptor of the input data. We evaluate the proposed model at image classification and action recognition problems and show how the model's classification performance improves as the amount of unlabeled data increases during training.
In this paper, we proposed a effective but extensible residual one-dimensional convolution neural network as base network, based on the this network, we proposed four subnets to explore the features of skeleton sequences from each aspect. Given a skeleton sequences, the spatial information are encoded into the skeleton joints coordinate in a frame and the temporal information are present by multiple frames. Limited by the skeleton sequence representations, two-dimensional convolution neural network cannot be used directly, we chose one-dimensional convolution layer as the basic layer. Each sub network could extract discriminative features from different aspects. Our first subnet is a two-stream network which could explore both temporal and spatial information. The second is a body-parted network, which could gain micro spatial features and macro temporal features. The third one is an attention network, the main contribution of which is to focus the key frames and feature channels which high related with the action classes in a skeleton sequence. One frame-difference network, as the last subnet, mainly processes the joints changes between the consecutive frames. Four subnets ensemble together by late fusion, the key problem of ensemble method is each subnet should have a certain performance and between the subnets, there are diversity existing. Each subnet shares a wellperformance basenet and differences between subnets guaranteed the diversity. Experimental results show that the ensemble network gets a state-of-the-art performance on three widely used datasets.
In video-based action recognition, viewpoint variations often pose major challenges because the same actions can appear different from different views. We use the complementary RGB and Depth information from the RGB-D cameras to address this problem. The proposed technique capitalizes on the spatio-temporal information available in the two data streams to the extract action features that are largely insensitive to the viewpoint variations. We use the RGB data to compute dense trajectories that are translated to viewpoint insensitive deep features under a non-linear knowledge transfer model. Similarly, the Depth stream is used to extract CNN-based view invariant features on which Fourier Temporal Pyramid is computed to incorporate the temporal information. The heterogeneous features from the two streams are combined and used as a dictionary to predict the label of the test samples. To that end, we propose a sparse-dense collaborative representation classification scheme that strikes a balance between the discriminative abilities of the dense and the sparse representations of the samples over the extracted heterogeneous dictionary.
In recent years, deep neural network approaches have naturally extended to the video domain, in their simplest case by aggregating per-frame classifications as a baseline for action recognition. A majority of the work in this area extends from the imaging domain, leading to visual-feature heavy approaches on temporal data. To address this issue we introduce "Let's Dance", a 1000 video dataset (and growing) comprised of 10 visually overlapping dance categories that require motion for their classification. We stress the important of human motion as a key distinguisher in our work given that, as we show in this work, visual information is not sufficient to classify motion-heavy categories. We compare our datasets' performance using imaging techniques with UCF-101 and demonstrate this inherent difficulty. We present a comparison of numerous state-of-the-art techniques on our dataset using three different representations (video, optical flow and multi-person pose data) in order to analyze these approaches. We discuss the motion parameterization of each of them and their value in learning to categorize online dance videos. Lastly, we release this dataset (and its three representations) for the research community to use.
We propose an action recognition framework using Gen- erative Adversarial Networks. Our model involves train- ing a deep convolutional generative adversarial network (DCGAN) using a large video activity dataset without la- bel information. Then we use the trained discriminator from the GAN model as an unsupervised pre-training step and fine-tune the trained discriminator model on a labeled dataset to recognize human activities. We determine good network architectural and hyperparameter settings for us- ing the discriminator from DCGAN as a trained model to learn useful representations for action recognition. Our semi-supervised framework using only appearance infor- mation achieves superior or comparable performance to the current state-of-the-art semi-supervised action recog- nition methods on two challenging video activity datasets: UCF101 and HMDB51.
Dynamics of human body skeletons convey significant information for human action recognition. Conventional approaches for modeling skeletons usually rely on hand-crafted parts or traversal rules, thus resulting in limited expressive power and difficulties of generalization. In this work, we propose a novel model of dynamic skeletons called Spatial-Temporal Graph Convolutional Networks (ST-GCN), which moves beyond the limitations of previous methods by automatically learning both the spatial and temporal patterns from data. This formulation not only leads to greater expressive power but also stronger generalization capability. On two large datasets, Kinetics and NTU-RGBD, it achieves substantial improvements over mainstream methods.
We present a generative framework for zero-shot action recognition where some of the possible action classes do not occur in the training data. Our approach is based on modeling each action class using a probability distribution whose parameters are functions of the attribute vector representing that action class. In particular, we assume that the distribution parameters for any action class in the visual space can be expressed as a linear combination of a set of basis vectors where the combination weights are given by the attributes of the action class. These basis vectors can be learned solely using labeled data from the known (i.e., previously seen) action classes, and can then be used to predict the parameters of the probability distributions of unseen action classes. We consider two settings: (1) Inductive setting, where we use only the labeled examples of the seen action classes to predict the unseen action class parameters; and (2) Transductive setting which further leverages unlabeled data from the unseen action classes. Our framework also naturally extends to few-shot action recognition where a few labeled examples from unseen classes are available. Our experiments on benchmark datasets (UCF101, HMDB51 and Olympic) show significant performance improvements as compared to various baselines, in both standard zero-shot (disjoint seen and unseen classes) and generalized zero-shot learning settings.
In this paper, we report on experiments with the use of local measures for depth motion for visual action recognition from MPEG encoded RGBD video sequences. We show that such measures can be combined with local space-time video descriptors for appearance to provide a computationally efficient method for recognition of actions. Fisher vectors are used for encoding and concatenating a depth descriptor with existing RGB local descriptors. We then employ a linear SVM for recognizing manipulation actions using such vectors. We evaluate the effectiveness of such measures by comparison to the state-of-the-art using two recent datasets for action recognition in kitchen environments.
We introduce the first benchmark for a new problem --- recognizing human action adverbs (HAA): "Adverbs Describing Human Actions" (ADHA). This is the first step for computer vision to change over from pattern recognition to real AI. We demonstrate some key features of ADHA: a semantically complete set of adverbs describing human actions, a set of common, describable human actions, and an exhaustive labeling of simultaneously emerging actions in each video. We commit an in-depth analysis on the implementation of current effective models in action recognition and image captioning on adverb recognition, and the results show that such methods are unsatisfactory. Moreover, we propose a novel three-stream hybrid model to deal the HAA problem, which achieves a better result.
Even with the recent advances in convolutional neural networks (CNN) in various visual recognition tasks, the state-of-the-art action recognition system still relies on hand crafted motion feature such as optical flow to achieve the best performance. We propose a multitask learning model ActionFlowNet to train a single stream network directly from raw pixels to jointly estimate optical flow while recognizing actions with convolutional neural networks, capturing both appearance and motion in a single model. We additionally provide insights to how the quality of the learned optical flow affects the action recognition. Our model significantly improves action recognition accuracy by a large margin 31% compared to state-of-the-art CNN-based action recognition models trained without external large scale data and additional optical flow input. Without pretraining on large external labeled datasets, our model, by well exploiting the motion information, achieves competitive recognition accuracy to the models trained with large labeled datasets such as ImageNet and Sport-1M.
Two-stream networks have been very successful for solving the problem of action detection. However, prior work using two-stream networks train both streams separately, which prevents the network from exploiting regularities between the two streams. Moreover, unlike the visual stream, the dominant forms of optical flow computation typically do not maximally exploit GPU parallelism. We present a real-time end-to-end trainable two-stream network for action detection. First, we integrate the optical flow computation in our framework by using Flownet2. Second, we apply early fusion for the two streams and train the whole pipeline jointly end-to-end. Finally, for better network initialization, we transfer from the task of action recognition to action detection by pre-training our framework using the recently released large-scale Kinetics dataset. Our experimental results show that training the pipeline jointly end-to-end with fine-tuning the optical flow for the objective of action detection improves detection performance significantly. Additionally, we observe an improvement when initializing with parameters pre-trained using Kinetics. Last, we show that by integrating the optical flow computation, our framework is more efficient, running at real-time speeds (up to 31 fps).
We propose a method for human activity recognition from RGB data that does not rely on any pose information during test time and does not explicitly calculate pose information internally. Instead, a visual attention module learns to predict glimpse sequences in each frame. These glimpses correspond to interest points in the scene that are relevant to the classified activities. No spatial coherence is forced on the glimpse locations, which gives the module liberty to explore different points at each frame and better optimize the process of scrutinizing visual information. Tracking and sequentially integrating this kind of unstructured data is a challenge, which we address by separating the set of glimpses from a set of recurrent tracking/recognition workers. These workers receive glimpses, jointly performing subsequent motion tracking and activity prediction. The glimpses are soft-assigned to the workers, optimizing coherence of the assignments in space, time and feature space using an external memory module. No hard decisions are taken, i.e. each glimpse point is assigned to all existing workers, albeit with different importance. Our methods outperform state-of-the-art methods on the largest human activity recognition dataset available to-date; NTU RGB+D Dataset, and on a smaller human action recognition dataset Northwestern-UCLA Multiview Action 3D Dataset. Our code is publicly available at https://github.com/fabienbaradel/glimpse_clouds.
Variations of human body skeletons may be considered as dynamic graphs, which are generic data representation for numerous real-world applications. In this paper, we propose a spatio-temporal graph convolution (STGC) approach for assembling the successes of local convolutional filtering and sequence learning ability of autoregressive moving average. To encode dynamic graphs, the constructed multi-scale local graph convolution filters, consisting of matrices of local receptive fields and signal mappings, are recursively performed on structured graph data of temporal and spatial domain. The proposed model is generic and principled as it can be generalized into other dynamic models. We theoretically prove the stability of STGC and provide an upper-bound of the signal transformation to be learnt. Further, the proposed recursive model can be stacked into a multi-layer architecture. To evaluate our model, we conduct extensive experiments on four benchmark skeleton-based action datasets, including the large-scale challenging NTU RGB+D. The experimental results demonstrate the effectiveness of our proposed model and the improvement over the state-of-the-art.
Although adversarial samples of deep neural networks (DNNs) have been intensively studied on static images, their extensions in videos are never explored. Compared with images, attacking a video needs to consider not only spatial cues but also temporal cues. Moreover, to improve the imperceptibility as well as reduce the computation cost, perturbations should be added on as fewer frames as possible, i.e., adversarial perturbations are temporally sparse. This further motivates the propagation of perturbations, which denotes that perturbations added on the current frame can transfer to the next frames via their temporal interactions. Thus, no (or few) extra perturbations are needed for these frames to misclassify them. To this end, we propose an l2,1-norm based optimization algorithm to compute the sparse adversarial perturbations for videos. We choose the action recognition as the targeted task, and networks with a CNN+RNN architecture as threat models to verify our method. Thanks to the propagation, we can compute perturbations on a shortened version video, and then adapt them to the long version video to fool DNNs. Experimental results on the UCF101 dataset demonstrate that even only one frame in a video is perturbed, the fooling rate can still reach 59.7%.
Classes in natural images tend to follow long tail distributions. This is problematic when there are insufficient training examples for rare classes. This effect is emphasized in compound classes, involving the conjunction of several concepts, such as those appearing in action-recognition datasets. In this paper, we propose to address this issue by learning how to utilize common visual concepts which are readily available. We detect the presence of prominent concepts in images and use them to infer the target labels instead of using visual features directly, combining tools from vision and natural-language processing. We validate our method on the recently introduced HICO dataset reaching a mAP of 31.54\% and on the Stanford-40 Actions dataset, where the proposed method outperforms that obtained by direct visual features, obtaining an accuracy 83.12\%. Moreover, the method provides for each class a semantically meaningful list of keywords and relevant image regions relating it to its constituent concepts.
Deep models are state-of-the-art for many vision tasks including video action recognition and video captioning. Models are trained to caption or classify activity in videos, but little is known about the evidence used to make such decisions. Grounding decisions made by deep networks has been studied in spatial visual content, giving more insight into model predictions for images. However, such studies are relatively lacking for models of spatiotemporal visual content - videos. In this work, we devise a formulation that simultaneously grounds evidence in space and time, in a single pass, using top-down saliency. We visualize the spatiotemporal cues that contribute to a deep model's classification/captioning output using the model's internal representation. Based on these spatiotemporal cues, we are able to localize segments within a video that correspond with a specific action, or phrase from a caption, without explicitly optimizing/training for these tasks.
In this paper, a novel video classification method is presented that aims to recognize different categories of third-person videos efficiently. Our motivation is to achieve a light model that could be trained with insufficient training data. With this intuition, the processing of the 3-dimensional video input is broken to 1D in temporal dimension on top of the 2D in spatial. The processes related to 2D spatial frames are being done by utilizing pre-trained networks with no training phase. The only step which involves training is to classify the 1D time series resulted from the description of the 2D signals. As a matter of fact, optical flow images are first calculated from consecutive frames and described by pre-trained CNN networks. Their dimension is then reduced using PCA. By stacking the description vectors beside each other, a multi-channel time series is created for each video. Each channel of the time series represents a specific feature and follows it over time. The main focus of the proposed method is to classify the obtained time series effectively. Towards this, the idea is to let the machine learn temporal features. This is done by training a multi-channel one dimensional Convolutional Neural Network (1D-CNN). The 1D-CNN learns the features along the only temporal dimension. Hence, the number of training parameters decreases significantly which would result in the trainability of the method on even smaller datasets. It is illustrated that the proposed method could reach the state-of-the-art results on two public datasets UCF11, jHMDB and competitive results on HMDB51.
Research in human action recognition has accelerated significantly since the introduction of powerful machine learning tools such as Convolutional Neural Networks (CNNs). However, effective and efficient methods for incorporation of temporal information into CNNs are still being actively explored in the recent literature. Motivated by the popular recurrent attention models in the research area of natural language processing, we propose the Attention-based Temporal Weighted CNN (ATW), which embeds a visual attention model into a temporal weighted multi-stream CNN. This attention model is simply implemented as temporal weighting yet it effectively boosts the recognition performance of video representations. Besides, each stream in the proposed ATW framework is capable of end-to-end training, with both network parameters and temporal weights optimized by stochastic gradient descent (SGD) with backpropagation. Our experiments show that the proposed attention mechanism contributes substantially to the performance gains with the more discriminative snippets by focusing on more relevant video segments.
This method introduces an efficient manner of learning action categories without the need of feature estimation. The approach starts from low-level values, in a similar style to the successful CNN methods. However, rather than extracting general image features, we learn to predict specific video representations from raw video data. The benefit of such an approach is that at the same computational expense it can predict 2 D video representations as well as 3 D ones, based on motion. The proposed model relies on discriminative Waldboost, which we enhance to a multiclass formulation for the purpose of learning video representations. The suitability of the proposed approach as well as its time efficiency are tested on the UCF11 action recognition dataset.
The computer vision community is currently focusing on solving action recognition problems in real videos, which contain thousands of samples with many challenges. In this process, Deep Convolutional Neural Networks (D-CNNs) have played a significant role in advancing the state-of-the-art in various vision-based action recognition systems. Recently, the introduction of residual connections in conjunction with a more traditional CNN model in a single architecture called Residual Network (ResNet) has shown impressive performance and great potential for image recognition tasks. In this paper, we investigate and apply deep ResNets for human action recognition using skeletal data provided by depth sensors. Firstly, the 3D coordinates of the human body joints carried in skeleton sequences are transformed into image-based representations and stored as RGB images. These color images are able to capture the spatial-temporal evolutions of 3D motions from skeleton sequences and can be efficiently learned by D-CNNs. We then propose a novel deep learning architecture based on ResNets to learn features from obtained color-based representations and classify them into action classes. The proposed method is evaluated on three challenging benchmark datasets including MSR Action 3D, KARD, and NTU-RGB+D datasets. Experimental results demonstrate that our method achieves state-of-the-art performance for all these benchmarks whilst requiring less computation resource. In particular, the proposed method surpasses previous approaches by a significant margin of 3.4% on MSR Action 3D dataset, 0.67% on KARD dataset, and 2.5% on NTU-RGB+D dataset.