Temporal action localization is an important yet challenging problem. Given a long, untrimmed video consisting of multiple action instances and complex background contents, we need not only to recognize their action categories, but also to localize the start time and end time of each instance. Many state-of-the-art systems use segment-level classifiers to select and rank proposal segments of pre-determined boundaries. However, a desirable model should move beyond segment-level and make dense predictions at a fine granularity in time to determine precise temporal boundaries. To this end, we design a novel Convolutional-De-Convolutional (CDC) network that places CDC filters on top of 3D ConvNets, which have been shown to be effective for abstracting action semantics but reduce the temporal length of the input data. The proposed CDC filter performs the required temporal upsampling and spatial downsampling operations simultaneously to predict actions at the frame-level granularity. It is unique in jointly modeling action semantics in space-time and fine-grained temporal dynamics. We train the CDC network in an end-to-end manner efficiently. Our model not only achieves superior performance in detecting actions in every frame, but also significantly boosts the precision of localizing temporal boundaries. Finally, the CDC network demonstrates a very high efficiency with the ability to process 500 frames per second on a single GPU server. We will update the camera-ready version and publish the source codes online soon.
Webly-supervised learning has recently emerged as an alternative paradigm to traditional supervised learning based on large-scale datasets with manual annotations. The key idea is that models such as CNNs can be learned from the noisy visual data available on the web. In this work we aim to exploit web data for video understanding tasks such as action recognition and detection. One of the main problems in webly-supervised learning is cleaning the noisy labeled data from the web. The state-of-the-art paradigm relies on training a first classifier on noisy data that is then used to clean the remaining dataset. Our key insight is that this procedure biases the second classifier towards samples that the first one understands. Here we train two independent CNNs, a RGB network on web images and video frames and a second network using temporal information from optical flow. We show that training the networks independently is vastly superior to selecting the frames for the flow classifier by using our RGB network. Moreover, we show benefits in enriching the training set with different data sources from heterogeneous public web databases. We demonstrate that our framework outperforms all other webly-supervised methods on two public benchmarks, UCF-101 and Thumos'14.
Skeleton-based human action recognition has attracted a lot of research attention during the past few years. Recent works attempted to utilize recurrent neural networks to model the temporal dependencies between the 3D positional configurations of human body joints for better analysis of human activities in the skeletal data. The proposed work extends this idea to spatial domain as well as temporal domain to better analyze the hidden sources of action-related information within the human skeleton sequences in both of these domains simultaneously. Based on the pictorial structure of Kinect's skeletal data, an effective tree-structure based traversal framework is also proposed. In order to deal with the noise in the skeletal data, a new gating mechanism within LSTM module is introduced, with which the network can learn the reliability of the sequential data and accordingly adjust the effect of the input data on the updating procedure of the long-term context representation stored in the unit's memory cell. Moreover, we introduce a novel multi-modal feature fusion strategy within the LSTM unit in this paper. The comprehensive experimental results on seven challenging benchmark datasets for human action recognition demonstrate the effectiveness of the proposed method.
In recent years a growing interest on action recognition is observed, including detection of fall accident for the elderly. However, despite many efforts undertaken, the existing technology is not widely used by elderly, mainly because of its flaws like low precision, large number of false alarms, inadequate privacy preserving during data acquisition and processing. This research work meets these expectations. The work is empirical and it is situated in the field of computer vision systems. The main part of the work situates itself in the area of action and behavior recognition. Efficient algorithms for fall detection were developed, tested and implemented using image sequences and wireless inertial sensor worn by a monitored person. A set of descriptors for depth maps has been elaborated to permit classification of pose as well as the action of a person. Experimental research was carried out based on the prepared data repository consisting of synchronized depth and accelerometric data. The study was carried out in the scenario with a static camera facing the scene and an active camera observing the scene from above. The experimental results showed that the developed algorithms for fall detection have high sensitivity and specificity. The algorithm were designed with regard to low computational demands and possibility to run on ARM platforms. Several experiments including person detection, tracking and fall detection in real-time were carried out to show efficiency and reliability of the proposed solutions.
We propose a novel method for temporally pooling frames in a video for the task of human action recognition. The method is motivated by the observation that there are only a small number of frames which, together, contain sufficient information to discriminate an action class present in a video, from the rest. The proposed method learns to pool such discriminative and informative frames, while discarding a majority of the non-informative frames in a single temporal scan of the video. Our algorithm does so by continuously predicting the discriminative importance of each video frame and subsequently pooling them in a deep learning framework. We show the effectiveness of our proposed pooling method on standard benchmarks where it consistently improves on baseline pooling methods, with both RGB and optical flow based Convolutional networks. Further, in combination with complementary video representations, we show results that are competitive with respect to the state-of-the-art results on two challenging and publicly available benchmark datasets.
A proper semantic representation for encoding side information is key to the success of zero-shot learning. In this paper, we explore two alternative semantic representations especially for zero-shot human action recognition: textual descriptions of human actions and deep features extracted from still images relevant to human actions. Such side information are accessible on Web with little cost, which paves a new way in gaining side information for large-scale zero-shot human action recognition. We investigate different encoding methods to generate semantic representations for human actions from such side information. Based on our zero-shot visual recognition method, we conducted experiments on UCF101 and HMDB51 to evaluate two proposed semantic representations . The results suggest that our proposed text- and image-based semantic representations outperform traditional attributes and word vectors considerably for zero-shot human action recognition. In particular, the image-based semantic representations yield the favourable performance even though the representation is extracted from a small number of images per class.
Action recognition is a fundamental problem in computer vision with a lot of potential applications such as video surveillance, human computer interaction, and robot learning. Given pre-segmented videos, the task is to recognize actions happening within videos. Historically, hand crafted video features were used to address the task of action recognition. With the success of Deep ConvNets as an image analysis method, a lot of extensions of standard ConvNets were purposed to process variable length video data. In this work, we propose a novel recurrent ConvNet architecture called recurrent residual networks to address the task of action recognition. The approach extends ResNet, a state of the art model for image classification. While the original formulation of ResNet aims at learning spatial residuals in its layers, we extend the approach by introducing recurrent connections that allow to learn a spatio-temporal residual. In contrast to fully recurrent networks, our temporal connections only allow a limited range of preceding frames to contribute to the output for the current frame, enabling efficient training and inference as well as limiting the temporal context to a reasonable local range around each frame. On a large-scale action recognition dataset, we show that our model improves over both, the standard ResNet architecture and a ResNet extended by a fully recurrent layer.
Recent methods based on 3D skeleton data have achieved outstanding performance due to its conciseness, robustness, and view-independent representation. With the development of deep learning, Convolutional Neural Networks (CNN) and Long Short Term Memory (LSTM)-based learning methods have achieved promising performance for action recognition. However, for CNN-based methods, it is inevitable to loss temporal information when a sequence is encoded into images. In order to capture as much spatial-temporal information as possible, LSTM and CNN are adopted to conduct effective recognition with later score fusion. In addition, experimental results show that the score fusion between CNN and LSTM performs better than that between LSTM and LSTM for the same feature. Our method achieved state-of-the-art results on NTU RGB+D datasets for 3D human action analysis. The proposed method achieved 87.40% in terms of accuracy and ranked $1^{st}$ place in Large Scale 3D Human Activity Analysis Challenge in Depth Videos.
Different from traditional action recognition based on video segments, online action recognition aims to recognize actions from unsegmented streams of data in a continuous manner. One way for online recognition is based on the evidence accumulation over time to make predictions from stream videos. This paper presents a fast yet effective method to recognize actions from stream of noisy skeleton data, and a novel weighted covariance descriptor is adopted to accumulate evidence. In particular, a fast incremental updating method for the weighted covariance descriptor is developed for accumulation of temporal information and online prediction. The weighted covariance descriptor takes the following principles into consideration: past frames have less contribution for recognition and recent and informative frames such as key frames contribute more to the recognition. The online recognition is achieved using a simple nearest neighbor search against a set of offline trained action models. Experimental results on MSC-12 Kinect Gesture dataset and our newly constructed online action recognition dataset have demonstrated the efficacy of the proposed method.
Landmark-based human action recognition in videos is a challenging task in computer vision. One key step is to design a generic approach that generates discriminative features for the spatial structure and temporal dynamics. To this end, we regard the evolving landmark data as a high-dimensional path and apply non-linear path signature techniques to provide an expressive, robust, non-linear, and interpretable representation for the sequential events. We do not extract signature features from the raw path, rather we propose path disintegrations and path transformations as preprocessing steps. Path disintegrations turn a high-dimensional path linearly into a collection of lower-dimensional paths; some of these paths are in pose space while others are defined over a multiscale collection of temporal intervals. Path transformations decorate the paths with additional coordinates in standard ways to allow the truncated signatures of transformed paths to expose additional features. For spatial representation, we apply the signature transform to vectorize the paths that arise out of pose disintegration, and for temporal representation, we apply it again to describe this evolving vectorization. Finally, all the features are collected together to constitute the input vector of a linear single-hidden-layer fully-connected network for classification. Experimental results on four datasets demonstrated that the proposed feature set with only a linear shallow network and Dropconnect is effective and achieves comparable state-of-the-art results to the advanced deep networks, and meanwhile, is capable of interpretation.
In this work we propose a novel neural network architecture for the problem of human action recognition in videos. The proposed architecture expresses the processing steps of classical Fisher vector approaches, that is dimensionality reduction by principal component analysis (PCA) projection, Gaussian mixture model (GMM) and Fisher vector descriptor extraction, as network layers. By contrast to other methods where these steps are performed consecutively and the corresponding parameters are learned in an unsupervised manner, having them defined as a single neural network allows us to refine the whole model discriminatively in an end to end fashion. Furthermore, we show that the proposed architecture can be used as a replacement for the fully connected layers in popular convolutional networks achieving a comparable classification performance, or even significantly surpassing the performance of similar architectures while reducing the total number of trainable parameters by a factor of 5. We show that our method achieves significant improvements in comparison to the classical chain.
Deep learning for human action recognition in videos is making significant progress, but is slowed down by its dependency on expensive manual labeling of large video collections. In this work, we investigate the generation of synthetic training data for action recognition, as it has recently shown promising results for a variety of other computer vision tasks. We propose an interpretable parametric generative model of human action videos that relies on procedural generation and other computer graphics techniques of modern game engines. We generate a diverse, realistic, and physically plausible dataset of human action videos, called PHAV for "Procedural Human Action Videos". It contains a total of 39,982 videos, with more than 1,000 examples for each action of 35 categories. Our approach is not limited to existing motion capture sequences, and we procedurally define 14 synthetic actions. We introduce a deep multi-task representation learning architecture to mix synthetic and real videos, even if the action categories differ. Our experiments on the UCF101 and HMDB51 benchmarks suggest that combining our large set of synthetic videos with small real-world datasets can boost recognition performance, significantly outperforming fine-tuning state-of-the-art unsupervised generative models of videos.
The production of sports highlight packages summarizing a game's most exciting moments is an essential task for broadcast media. Yet, it requires labor-intensive video editing. We propose a novel approach for auto-curating sports highlights, and use it to create a real-world system for the editorial aid of golf highlight reels. Our method fuses information from the players' reactions (action recognition such as high-fives and fist pumps), spectators (crowd cheering), and commentator (tone of the voice and word analysis) to determine the most interesting moments of a game. We accurately identify the start and end frames of key shot highlights with additional metadata, such as the player's name and the hole number, allowing personalized content summarization and retrieval. In addition, we introduce new techniques for learning our classifiers with reduced manual training data annotation by exploiting the correlation of different modalities. Our work has been demonstrated at a major golf tournament, successfully extracting highlights from live video streams over four consecutive days.
In this notebook paper, we describe our approach in the submission to the temporal action proposal (task 3) and temporal action localization (task 4) of ActivityNet Challenge hosted at CVPR 2017. Since the accuracy in action classification task is already very high (nearly 90% in ActivityNet dataset), we believe that the main bottleneck for temporal action localization is the quality of action proposals. Therefore, we mainly focus on the temporal action proposal task and propose a new proposal model based on temporal convolutional network. Our approach achieves the state-of-the-art performances on both temporal action proposal task and temporal action localization task.
Most popular deep models for action recognition split video sequences into short sub-sequences consisting of a few frames; frame-based features are then pooled for recognizing the activity. Usually, this pooling step discards the temporal order of the frames, which could otherwise be used for better recognition. Towards this end, we propose a novel pooling method, generalized rank pooling (GRP), that takes as input, features from the intermediate layers of a CNN that is trained on tiny sub-sequences, and produces as output the parameters of a subspace which (i) provides a low-rank approximation to the features and (ii) preserves their temporal order. We propose to use these parameters as a compact representation for the video sequence, which is then used in a classification setup. We formulate an objective for computing this subspace as a Riemannian optimization problem on the Grassmann manifold, and propose an efficient conjugate gradient scheme for solving it. Experiments on several activity recognition datasets show that our scheme leads to state-of-the-art performance.
Actions are more than just movements and trajectories: we cook to eat and we hold a cup to drink from it. A thorough understanding of videos requires going beyond appearance modeling and necessitates reasoning about the sequence of activities, as well as the higher-level constructs such as intentions. But how do we model and reason about these? We propose a fully-connected temporal CRF model for reasoning over various aspects of activities that includes objects, actions, and intentions, where the potentials are predicted by a deep network. End-to-end training of such structured models is a challenging endeavor: For inference and learning we need to construct mini-batches consisting of whole videos, leading to mini-batches with only a few videos. This causes high-correlation between data points leading to breakdown of the backprop algorithm. To address this challenge, we present an asynchronous variational inference method that allows efficient end-to-end training. Our method achieves a classification mAP of 22.4% on the Charades benchmark, outperforming the state-of-the-art (17.2% mAP), and offers equal gains on the task of temporal localization.
Discriminative clustering has been successfully applied to a number of weakly-supervised learning tasks. Such applications include person and action recognition, text-to-video alignment, object co-segmentation and colocalization in videos and images. One drawback of discriminative clustering, however, is its limited scalability. We address this issue and propose an online optimization algorithm based on the Block-Coordinate Frank-Wolfe algorithm. We apply the proposed method to the problem of weakly supervised learning of actions and actors from movies together with corresponding movie scripts. The scaling up of the learning problem to 66 feature length movies enables us to significantly improve weakly supervised action recognition.
Pose detection is one of the fundamental steps for the recognition of human actions. In this paper we propose a novel trainable detector for recognizing human poses based on the analysis of the skeleton. The main idea is that a skeleton pose can be described by the spatial arrangements of its joints. Starting from this consideration, we propose a trainable pose detector, that can be configured on a prototype skeleton in an automatic configuration process. The result of the configuration is a model of the position of the joints in the concerned skeleton. In the application phase, the joint positions contained in the model are compared with the ones of their homologous joints in the skeleton under test. The similarity of two skeletons is computed as a combination of the position scores achieved by homologous joints. In this paper we describe an action classification method based on the use of the proposed trainable detectors to extract features from the skeletons. We performed experiments on the publicly available MSDRA data set and the achieved results confirm the effectiveness of the proposed approach.
We present an unsupervised representation learning approach using videos without semantic labels. We leverage the temporal coherence as a supervisory signal by formulating representation learning as a sequence sorting task. We take temporally shuffled frames (i.e., in non-chronological order) as inputs and train a convolutional neural network to sort the shuffled sequences. Similar to comparison-based sorting algorithms, we propose to extract features from all frame pairs and aggregate them to predict the correct order. As sorting shuffled image sequence requires an understanding of the statistical temporal structure of images, training with such a proxy task allows us to learn rich and generalizable visual representation. We validate the effectiveness of the learned representation using our method as pre-training on high-level recognition problems. The experimental results show that our method compares favorably against state-of-the-art methods on action recognition, image classification and object detection tasks.
Human action recognition from skeletal data is a hot research topic and important in many open domain applications of computer vision, thanks to recently introduced 3D sensors. In the literature, naive methods simply transfer off-the-shelf techniques from video to the skeletal representation. However, the current state-of-the-art is contended between to different paradigms: kernel-based methods and feature learning with (recurrent) neural networks. Both approaches show strong performances, yet they exhibit heavy, but complementary, drawbacks. Motivated by this fact, our work aims at combining together the best of the two paradigms, by proposing an approach where a shallow network is fed with a covariance representation. Our intuition is that, as long as the dynamics is effectively modeled, there is no need for the classification network to be deep nor recurrent in order to score favorably. We validate this hypothesis in a broad experimental analysis over 6 publicly available datasets.