This paper proposes a novel human action recognition using the decision-level fusion of both skeleton and depth sequence. Firstly, a state-of-the-art descriptor RBPL, relative body part locations, is adopted to represent skeleton. But the original RBPL employs all the available joints, which may introduce redundancy or noise. This paper proposes an adaptive optimal joint selection model based on the distance traveled by joints before RBPL for each different action, which can reduce redundant joints. Then we use dynamic time warping to handle temporal misalignment and adopt KELM, kernel-based extreme learning machine, for action recognition. Secondly, an efficient feature descriptor DMM-disLBP, depth motion maps-based discriminative local binary patterns, is constructed to describe depth sequences, and KELM is also used for classification. Finally, we present an effective decision fusion for action recognition based on the maximum sum of decision values from skeleton and depth maps. Comparing with the baseline methods, we improve the performance using either skeleton or depth information, and achieve the state-of-the-art average recognition accuracy on the public dataset MSR Action3D using proposed fusing strategy.
A great number of computer vision publications have focused on distinguishing between human action recognition and classification rather than the intensity of actions performed. Indexing the intensity which determines the performance of human actions is a challenging task due to the uncertainty and information deficiency that exists in the video inputs. To remedy this uncertainty, in this paper we coupled fuzzy logic rules with the neural-based action recognition model to rate the intensity of a human action as intense or mild. In our approach, we used a Spatio-Temporal LSTM to generate the weights of the fuzzy-logic model, and then demonstrate through experiments that indexing of the action intensity is possible. We analyzed the integrated model by applying it to videos of human actions with different action intensities and were able to achieve an accuracy of 89.16% on our intensity indexing generated dataset. The integrated model demonstrates the ability of a neuro-fuzzy inference module to effectively estimate the intensity index of human actions.
Temporal action localization is an important step towards video understanding. Most current action localization methods depend on untrimmed videos with full temporal annotations of action instances. However, it is expensive and time-consuming to annotate both action labels and temporal boundaries of videos. To this end, we propose a weakly supervised temporal action localization method that only requires video-level action instances as supervision during training. We propose a classification module to generate action labels for each segment in the video, and a deep metric learning module to learn the similarity between different action instances. We jointly optimize a balanced binary cross-entropy loss and a metric loss using a standard backpropagation algorithm. Extensive experiments demonstrate the effectiveness of both of these components in temporal localization. We evaluate our algorithm on two challenging untrimmed video datasets: THUMOS14 and ActivityNet1.2. Our approach improves the current state-of-the-art result for THUMOS14 by 6.5% mAP at IoU threshold 0.5, and achieves competitive performance for ActivityNet1.2.
Skeleton-based human action recognition has become an active research area in recent years. The key to this task is to fully explore both spatial and temporal features. Recently, GCN-based methods modeling the human body skeletons as spatial-temporal graphs, have achieved remarkable performances. However, most GCN-based methods use a fixed adjacency matrix defined by the dataset, which can only capture the structural information provided by joints directly connected through bones and ignore the dependencies between distant joints that are not connected. In addition, such a fixed adjacency matrix used in all layers leads to the network failing to extract multi-level semantic features. In this paper we propose a pseudo graph convolutional network with temporal and channel-wise attention (PGCN-TCA) to solve this problem. The fixed normalized adjacent matrix is substituted with a learnable matrix. In this way, the matrix can learn the dependencies between connected joints and joints that are not physically connected. At the same time, learnable matrices in different layers can help the network capture multi-level features in spatial domain. Moreover, Since frames and input channels that contain outstanding characteristics play significant roles in distinguishing the action from others, we propose a mixed temporal and channel-wise attention. Our method achieves comparable performances to state-of-the-art methods on NTU-RGB+D and HDM05 datasets.
The robust recognition and assessment of human actions are crucial in human-robot interaction (HRI) domains. While state-of-the-art models of action perception show remarkable results in large-scale action datasets, they mostly lack the flexibility, robustness, and scalability needed to operate in natural HRI scenarios which require the continuous acquisition of sensory information as well as the classification or assessment of human body patterns in real time. In this chapter, I introduce a set of hierarchical models for the learning and recognition of actions from depth maps and RGB images through the use of neural network self-organization. A particularity of these models is the use of growing self-organizing networks that quickly adapt to non-stationary distributions and implement dedicated mechanisms for continual learning from temporally correlated input.
In this work, we address the problem of learning an ensemble of specialist networks using multimodal data, while considering the realistic and challenging scenario of possible missing modalities at test time. Our goal is to leverage the complementary information of multiple modalities to the benefit of the ensemble and each individual network. We introduce a novel Distillation Multiple Choice Learning framework for multimodal data, where different modality networks learn in a cooperative setting from scratch, strengthening one another. The modality networks learned using our method achieve significantly higher accuracy than if trained separately, due to the guidance of other modalities. We evaluate this approach on three video action recognition benchmark datasets. We obtain state-of-the-art results in comparison to other approaches that work with missing modalities at test time.
Skeleton-based action recognition has recently attracted a lot of attention. Researchers are coming up with new approaches for extracting spatio-temporal relations and making considerable progress on large-scale skeleton-based datasets. Most of the architectures being proposed are based upon recurrent neural networks (RNNs), convolutional neural networks (CNNs) and graph-based CNNs. When it comes to skeleton-based action recognition, the importance of long term contextual information is central which is not captured by the current architectures. In order to come up with a better representation and capturing of long term spatio-temporal relationships, we propose three variants of Self-Attention Network (SAN), namely, SAN-V1, SAN-V2 and SAN-V3. Our SAN variants has the impressive capability of extracting high-level semantics by capturing long-range correlations. We have also integrated the Temporal Segment Network (TSN) with our SAN variants which resulted in improved overall performance. Different configurations of Self-Attention Network (SAN) variants and Temporal Segment Network (TSN) are explored with extensive experiments. Our chosen configuration outperforms state-of-the-art Top-1 and Top-5 by 4.4% and 7.9% respectively on Kinetics and shows consistently better performance than state-of-the-art methods on NTU RGB+D.
Recent methods for video action recognition have reached outstanding performances on existing benchmarks. However, they tend to leverage context such as scenes or objects instead of focusing on understanding the human action itself. For instance, a tennis field leads to the prediction playing tennis irrespectively of the actions performed in the video. In contrast, humans have a more complete understanding of actions and can recognize them without context. The best example of out-of-context actions are mimes, that people can typically recognize despite missing relevant objects and scenes. In this paper, we propose to benchmark action recognition methods in such absence of context and introduce a novel dataset, Mimetics, consisting of mimed actions for a subset of 50 classes from the Kinetics benchmark. Our experiments show that (a) state-of-the-art 3D convolutional neural networks obtain disappointing results on such videos, highlighting the lack of true understanding of the human actions and (b) models leveraging body language via human pose are less prone to context biases. In particular, we show that applying a shallow neural network with a single temporal convolution over body pose features transferred to the action recognition problem performs surprisingly well compared to 3D action recognition methods.
Graph convolutional networks (GCNs), which generalize CNNs to more generic non-Euclidean structures, have achieved remarkable performance for skeleton-based action recognition. However, there still exist several issues in the previous GCN-based models. First, the topology of the graph is set heuristically and fixed over all the model layers and input data. This may not be suitable for the hierarchy of the GCN model and the diversity of the data in action recognition tasks. Second, the second-order information of the skeleton data, i.e., the length and orientation of the bones, is rarely investigated, which is naturally more informative and discriminative for the human action recognition. In this work, we propose a novel multi-stream attention-enhanced adaptive graph convolutional neural network (MS-AAGCN) for skeleton-based action recognition. The graph topology in our model can be either uniformly or individually learned based on the input data in an end-to-end manner. This data-driven approach increases the flexibility of the model for graph construction and brings more generality to adapt to various data samples. Besides, the proposed adaptive graph convolutional layer is further enhanced by a spatial-temporal-channel attention module, which helps the model pay more attention to important joints, frames and features. Moreover, the information of both the joints and bones, together with their motion information, are simultaneously modeled in a multi-stream framework, which shows notable improvement for the recognition accuracy. Extensive experiments on the two large-scale datasets, NTU-RGBD and Kinetics-Skeleton, demonstrate that the performance of our model exceeds the state-of-the-art with a significant margin.
Soccer videos can serve as a perfect research object for video understanding because soccer games are played under well-defined rules while complex and intriguing enough for researchers to study. In this paper, we propose a new soccer video database named SoccerDB, comprising 171,191 video segments from 346 high-quality soccer games. The database contains 702,096 bounding boxes, 37,709 essential event labels with time boundary and 17,115 highlight annotations for object detection, action recognition, temporal action localization, and highlight detection tasks. To our knowledge, it is the largest database for comprehensive sports video understanding on various aspects. We further survey a collection of strong baselines on SoccerDB, which have demonstrated state-of-the-art performances on independent tasks. Our evaluation suggests that we can benefit significantly when jointly considering the inner correlations among those tasks. We believe the release of SoccerDB will tremendously advance researches around comprehensive video understanding. {\itshape Our dataset and code published on https://github.com/newsdata/SoccerDB.}
Two-stream networks have achieved great success in video recognition. A two-stream network combines a spatial stream of RGB frames and a temporal stream of Optical Flow to make predictions. However, the temporal redundancy of RGB frames as well as the high-cost of optical flow computation creates challenges for both the performance and efficiency. Recent works instead use modern compressed video modalities as an alternative to the RGB spatial stream and improve the inference speed by orders of magnitudes. Previous works create one stream for each modality which are combined with an additional temporal stream through late fusion. This is redundant since some modalities like motion vectors already contain temporal information. Based on this observation, we propose a compressed domain two-stream network IP TSN for compressed video recognition, where the two streams are represented by the two types of frames (I and P frames) in compressed videos, without needing a separate temporal stream. With this goal, we propose to fully exploit the motion information of P-stream through generalized distillation from optical flow, which largely improves the efficiency and accuracy. Our P-stream runs 60 times faster than using optical flow while achieving higher accuracy. Our full IP TSN, evaluated over public action recognition benchmarks (UCF101, HMDB51 and a subset of Kinetics), outperforms other compressed domain methods by large margins while improving the total inference speed by 20%.
Action Detection is a complex task that aims to detect and classify human actions in video clips. Typically, it has been addressed by processing fine-grained features extracted from a video classification backbone. Recently, thanks to the robustness of object and people detectors, a deeper focus has been added on relationship modelling. Following this line, we propose a graph-based framework to learn high-level interactions between people and objects, in both space and time. In our formulation, spatio-temporal relationships are learned through self-attention on a multi-layer graph structure which can connect entities from consecutive clips, thus considering long-range spatial and temporal dependencies. The proposed module is backbone independent by design and does not require end-to-end training. Extensive experiments are conducted on the AVA dataset, where our model demonstrates state-of-the-art results and consistent improvements over baselines built with different backbones. Code is publicly available at https://github.com/aimagelab/STAGE_action_detection.
Recognition of human actions and associated interactions with objects and the environment is an important problem in computer vision due to its potential applications in a variety of domains. The most versatile methods can generalize to various environments and deal with cluttered backgrounds, occlusions, and viewpoint variations. Among them, methods based on graph convolutional networks that extract features from the skeleton have demonstrated promising performance. In this paper, we propose a novel Spatio-Temporal Pyramid Graph Convolutional Network (ST-PGN) for online action recognition for ergonomic risk assessment that enables the use of features from all levels of the skeleton feature hierarchy. The proposed algorithm outperforms state-of-art action recognition algorithms tested on two public benchmark datasets typically used for postural assessment (TUM and UW-IOM). We also introduce a pipeline to enhance postural assessment methods with online action recognition techniques. Finally, the proposed algorithm is integrated with a traditional ergonomic risk index (REBA) to demonstrate the potential value for assessment of musculoskeletal disorders in occupational safety.
Human action Recognition for unknown views is a challenging task. We propose a view-invariant deep human action recognition framework, which is a novel integration of two important action cues: motion and shape temporal dynamics (STD). The motion stream encapsulates the motion content of action as RGB Dynamic Images (RGB-DIs) which are processed by the fine-tuned InceptionV3 model. The STD stream learns long-term view-invariant shape dynamics of action using human pose model (HPM) based view-invariant features mined from structural similarity index matrix (SSIM) based key depth human pose frames. To predict the score of the test sample, three types of late fusion (maximum, average and product) techniques are applied on individual stream scores. To validate the performance of the proposed novel framework the experiments are performed using both cross subject and cross-view validation schemes on three publically available benchmarks- NUCLA multi-view dataset, UWA3D-II Activity dataset and NTU RGB-D Activity dataset. Our algorithm outperforms with existing state-of-the-arts significantly that is reported in terms of accuracy, receiver operating characteristic (ROC) curve and area under the curve (AUC).
In this report, we introduce the Winner method for HACS Temporal Action Localization Challenge 2019. Temporal action localization is challenging since a target proposal may be related to several other candidate proposals in an untrimmed video. Existing methods cannot tackle this challenge well since temporal proposals are considered individually and their temporal dependencies are neglected. To address this issue, we propose sparse 2D temporal adjacent networks to model the temporal relationship between candidate proposals. This method is built upon the recent proposed 2D-TAN approach. The sampling strategy in 2D-TAN introduces the unbalanced context problem, where short proposals can perceive more context than long proposals. Therefore, we further propose a Sparse 2D Temporal Adjacent Network (S-2D-TAN). It is capable of involving more context information for long proposals and further learning discriminative features from them. By combining our S-2D-TAN with a simple action classifier, our method achieves a mAP of 23.49 on the test set, which win the first place in the HACS challenge.
Although synthetic training data has been shown to be beneficial for tasks such as human pose estimation, its use for RGB human action recognition is relatively unexplored. Our goal in this work is to answer the question whether synthetic humans can improve the performance of human action recognition, with a particular focus on generalization to unseen viewpoints. We make use of the recent advances in monocular 3D human body reconstruction from real action sequences to automatically render synthetic training videos for the action labels. We make the following contributions: (i) we investigate the extent of variations and augmentations that are beneficial to improving performance at new viewpoints. We consider changes in body shape and clothing for individuals, as well as more action relevant augmentations such as non-uniform frame sampling, and interpolating between the motion of individuals performing the same action; (ii) We introduce a new data generation methodology, SURREACT, that allows training of spatio-temporal CNNs for action classification; (iii) We substantially improve the state-of-the-art action recognition performance on the NTU RGB+D and UESTC standard human action multi-view benchmarks; Finally, (iv) we extend the augmentation approach to in-the-wild videos from a subset of the Kinetics dataset to investigate the case when only one-shot training data is available, and demonstrate improvements in this case as well.
Despite the rapid growth in datasets for video activity, stable robust activity recognition with neural networks remains challenging. This is in large part due to the explosion of possible variation in video -- including lighting changes, object variation, movement variation, and changes in surrounding context. An alternative is to make use of simulation data, where all of these factors can be artificially controlled. In this paper, we propose the Randomized Simulation as Augmentation (RSA) framework which augments real-world training data with synthetic data to improve the robustness of action recognition networks. We generate large-scale synthetic datasets with randomized nuisance factors. We show that training with such extra data, when appropriately constrained, can significantly improve the performance of the state-of-the-art I3D networks or, conversely, reduce the number of labeled real videos needed to achieve good performance. Experiments on two real-world datasets NTU RGB+D and VIRAT demonstrate the effectiveness of our method.
Current state-of-the-art models for video action recognition are mostly based on expensive 3D ConvNets. This results in a need for large GPU clusters to train and evaluate such architectures. To address this problem, we present a lightweight and memory-friendly architecture for action recognition that performs on par with or better than current architectures by using only a fraction of resources. The proposed architecture is based on a combination of a deep subnet operating on low-resolution frames with a compact subnet operating on high-resolution frames, allowing for high efficiency and accuracy at the same time. We demonstrate that our approach achieves a reduction by $3\sim4$ times in FLOPs and $\sim2$ times in memory usage compared to the baseline. This enables training deeper models with more input frames under the same computational budget. To further obviate the need for large-scale 3D convolutions, a temporal aggregation module is proposed to model temporal dependencies in a video at very small additional computational costs. Our models achieve strong performance on several action recognition benchmarks including Kinetics, Something-Something and Moments-in-time. The code and models are available at https://github.com/IBM/bLVNet-TAM.
There exist a wide range of intra class variations of the same actions and inter class similarity among the actions, at the same time, which makes the action recognition in videos very challenging. In this paper, we present a novel skeleton-based part-wise Spatiotemporal CNN RIAC Network-based 3D human action recognition framework to visualise the action dynamics in part wise manner and utilise each part for action recognition by applying weighted late fusion mechanism. Part wise skeleton based motion dynamics helps to highlight local features of the skeleton which is performed by partitioning the complete skeleton in five parts such as Head to Spine, Left Leg, Right Leg, Left Hand, Right Hand. The RIAFNet architecture is greatly inspired by the InceptionV4 architecture which unified the ResNet and Inception based Spatio-temporal feature representation concept and achieving the highest top-1 accuracy till date. To extract and learn salient features for action recognition, attention driven residues are used which enhance the performance of residual components for effective 3D skeleton-based Spatio-temporal action representation. The robustness of the proposed framework is evaluated by performing extensive experiments on three challenging datasets such as UT Kinect Action 3D, Florence 3D action Dataset, and MSR Daily Action3D datasets, which consistently demonstrate the superiority of our method
Pose-based action recognition has drawn considerable attention recently. Existing methods exploit the joint positions to extract the body-part features from the activation map of the convolutional networks to assist human action recognition. However, these features are simply concatenated or max-pooled in previous works. The structured correlations among the body parts, which are essential for understanding complex human actions, are not fully exploited. To address the problem, we propose a pose-based graph convolutional network (PGCN), which encodes the body-part features into a human-based spatiotemporal graph, and explicitly models their correlations with a novel light-weight adaptive graph convolutional module to produce a highly discriminative representation for human action recognition. Besides, we discover that the backbone network tends to identify patterns from the most discriminative areas of the input regardless of the others. Thus the features pooled by the joint positions from other areas are less informative, which consequently hampers the performance of the followed aggregation process for recognizing actions. To alleviate this issue, we introduce a simple intermediate dense supervision mechanism for the backbone network, which adequately addresses the problem with no extra computation cost during inference. We evaluate the proposed approach on three popular benchmarks for pose-based action recognition tasks, i.e., Sub-JHMDB, PennAction and NTU-RGBD, where our approach significantly outperforms state-of-the-arts without the bells and whistles.