Contributive Representation-Based Reconstruction for Online 3D Action Recognition

2020 ◽  
pp. 2150005
Author(s):  
Mohsen Tabejamaat ◽  
Hoda Mohammadzade
Keyword(s):  
Action Recognition ◽  
Experimental Results ◽  
Initial Part ◽  
Optimal Classifier ◽  
Training Sequences ◽  
Action Sequences ◽  
Increasing Trend ◽  
Sequence Representation ◽  
Existing Frames ◽  
Query Sample

Recent years have seen an increasing trend in developing 3D action recognition methods. However, despite the advances, existing models still suffer from some major drawbacks including the lack of any provision for recognizing action sequences with some missing frames. This significantly hampers the applicability of these methods for online scenarios, where only an initial part of sequences are already provided. In this paper, we introduce a novel sequence-to-sequence representation-based algorithm in which a query sample is characterized using a collaborative frame representation of all the training sequences. This way, an optimal classifier is tailored for the existing frames of each query sample, making the model robust to the effect of missing frames in sequences (e.g. in online scenarios). Moreover, due to the collaborative nature of the representation, it implicitly handles the problem of varying styles during the course of activities. Experimental results on three publicly available databases, UTKinect, TST fall, and UTD-MHAD, respectively, show 95.48%, 90.91%, and 91.67% accuracy when using the beginning 75% portion of query sequences and 84.42%, 60.98%, and 87.27% accuracy for their initial 50%.

scholarly journals Action Recognition Using Close-Up of Maximum Activation and ETRI-Activity3D LivingLab Dataset

Sensors ◽  
2021 ◽  
Vol 21 (20) ◽  
pp. 6774
Author(s):  
Doyoung Kim ◽  
Inwoong Lee ◽  
Dohyung Kim ◽  
Sanghoon Lee
Keyword(s):  
Action Recognition ◽  
High Availability ◽  
Video Data ◽  
Experimental Results ◽  
Great Performance ◽  
Action Sequences ◽  
Maximum Activation ◽  
Rich Data ◽  
Video Input ◽  
Dataset Shift

The development of action recognition models has shown great performance on various video datasets. Nevertheless, because there is no rich data on target actions in existing datasets, it is insufficient to perform action recognition applications required by industries. To satisfy this requirement, datasets composed of target actions with high availability have been created, but it is difficult to capture various characteristics in actual environments because video data are generated in a specific environment. In this paper, we introduce a new ETRI-Activity3D-LivingLab dataset, which provides action sequences in actual environments and helps to handle a network generalization issue due to the dataset shift. When the action recognition model is trained on the ETRI-Activity3D and KIST SynADL datasets and evaluated on the ETRI-Activity3D-LivingLab dataset, the performance can be severely degraded because the datasets were captured in different environments domains. To reduce this dataset shift between training and testing datasets, we propose a close-up of maximum activation, which magnifies the most activated part of a video input in detail. In addition, we present various experimental results and analysis that show the dataset shift and demonstrate the effectiveness of the proposed method.

scholarly journals Automatic Detection of Discrimination Actions from Social Images

Electronics ◽  
2021 ◽  
Vol 10 (3) ◽  
pp. 325
Author(s):  
Zhihao Wu ◽  
Baopeng Zhang ◽  
Tianchen Zhou ◽  
Yan Li ◽  
Jianping Fan
Keyword(s):  
Action Recognition ◽  
State Of The Art ◽  
Automatic Detection ◽  
Experimental Results ◽  
Practical Approach ◽  
Detection And Identification ◽  
Art Methods ◽  
Image Set ◽  
Social Images ◽  
Relationship Identification

In this paper, we developed a practical approach for automatic detection of discrimination actions from social images. Firstly, an image set is established, in which various discrimination actions and relations are manually labeled. To the best of our knowledge, this is the first work to create a dataset for discrimination action recognition and relationship identification. Secondly, a practical approach is developed to achieve automatic detection and identification of discrimination actions and relationships from social images. Thirdly, the task of relationship identification is seamlessly integrated with the task of discrimination action recognition into one single network called the Co-operative Visual Translation Embedding++ network (CVTransE++). We also compared our proposed method with numerous state-of-the-art methods, and our experimental results demonstrated that our proposed methods can significantly outperform state-of-the-art approaches.

Preliminary Experiments for Evaluating 3-D Effects on Cracks in Frozen Stress Models

2003 ◽  
Author(s):  
C. W. Smith ◽  
J. D. Hansen ◽  
C. T. Liu
Keyword(s):  
Experimental Results ◽  
Design Rationale ◽  
Initial Part ◽  
Photoelastic Method ◽  
Solid Rocket Motors ◽  
Rocket Motors ◽  
Background Data ◽  
Solid Rocket ◽  
Stress Models ◽  
Critical Locations

Information gleaned from applying the frozen stress photoelastic method to cracks emanating from critical locations around the fin tips in models of solid rocket motors is reviewed and assessed together with new experimental results. The studies are the initial part of a program developed to contribute background data for consideration in modifying current motor grain design rationale.

scholarly journals Online Strategy Clustering Based on Action Sequences in RoboCupSoccer Small Size League

Robotics ◽  
2019 ◽  
Vol 8 (3) ◽  
pp. 58
Author(s):  
Yusuke Adachi ◽  
Masahide Ito ◽  
Tadashi Naruse
Keyword(s):  
Experimental Results ◽  
Clustering Method ◽  
Learning Problem ◽  
Data Set ◽  
Geometric Data ◽  
Action Sequences ◽  
Novel Method ◽  
Online Strategy

This paper addresses a strategy learning problem in the RoboCupSoccer Small Size League (SSL). We propose a novel method based on action sequences to cluster an opponent’s strategies online. Our proposed method is composed of the following three steps: (1) extracting typical actions from geometric data to make action sequences, (2) calculating the dissimilarity of the sequences, and (3) clustering the sequences by using the dissimilarity. This method can reduce the amount of data used in the clustering process; handling action sequences instead of geometric data as data-set makes it easier to search actions. As a result, the proposed clustering method is online feasible and also is applicable to countering an opponent’s strategy. The effectiveness of the proposed method was validated by experimental results.

Human Action Recognition Algorithm Based on Key Posture

2013 ◽  
Vol 631-632 ◽  
pp. 1303-1308
Author(s):  
He Jin Yuan
Keyword(s):  
Action Recognition ◽  
Clustering Algorithm ◽  
Recognition Accuracy ◽  
Human Action Recognition ◽  
Human Action ◽  
Recognition Algorithm ◽  
Training Samples ◽  
Action Sequences

A novel human action recognition algorithm based on key posture is proposed in this paper. In the method, the mesh features of each image in human action sequences are firstly calculated; then the key postures of the human mesh features are generated through k-medoids clustering algorithm; and the motion sequences are thus represented as vectors of key postures. The component of the vector is the occurrence number of the corresponding posture included in the action. For human action recognition, the observed action is firstly changed into key posture vector; then the correlevant coefficients to the training samples are calculated and the action which best matches the observed sequence is chosen as the final category. The experiments on Weizmann dataset demonstrate that our method is effective for human action recognition. The average recognition accuracy can exceed 90%.

Research on Applied Technology in Human Action Recognition Based on Skeleton Information

2013 ◽  
Vol 859 ◽  
pp. 498-502 ◽  
Author(s):  
Zhi Qiang Wei ◽  
Ji An Wu ◽  
Xi Wang
Keyword(s):  
Data Analysis ◽  
Action Recognition ◽  
Recognition Rate ◽  
Human Action Recognition ◽  
Human Action ◽  
Experimental Results ◽  
Human Actions ◽  
Applied Technology ◽  
Feature Sequence ◽  
Human Actions Recognition

In order to realize the identification of human daily actions, a method of identifying human daily actions is realized in this paper, which transforms this problem into converting human action recognition into analyzing feature sequence. Then the feature sequence combined with improved LCS algorithm could realize the human actions recognition. Data analysis and experimental results show the recognition rate of this method is high and speed is fast, and this applied technology will have broad prospects.

An Adaptive Stacked Denoising Auto-Encoder Architecture for Human Action Recognition

2014 ◽  
Vol 631-632 ◽  
pp. 403-409 ◽  
Author(s):  
Dao Xi Wu ◽  
Wei Pan ◽  
Li Dong Xie ◽  
Chao Xi Huang
Keyword(s):  
Neural Network ◽  
Action Recognition ◽  
Adaptive Learning ◽  
Human Action Recognition ◽  
Human Action ◽  
Learning Rate ◽  
Experimental Results ◽  
Classification Methods ◽  
Initial Value ◽  
Human Actions

In this paper, a stacked denoising auto-encoder architecture method with adaptive learning rate for action recognition based on skeleton features of human is presented. Firstly a Kinect is used for capturing the skeleton images and extracting skeleton features. Then an adaptive stacked denoising auto-encoder with three hidden layers is constructed for unsupervised pre-training. So the trained weights are achieved. Finally, a neural network is constructed for action recognition, in which the trained weights are used as the initial value, covering the random value. Based on the experimental results from the Kinect dataset of human actions sampled in experiments, it is clear to see that our method possesses the better robustness and accuracy, compared with the classic classification methods.

scholarly journals Monitoring of Assembly Process Using Deep Learning Technology

Sensors ◽  
2020 ◽  
Vol 20 (15) ◽  
pp. 4208
Author(s):  
Chengjun Chen ◽  
Chunlin Zhang ◽  
Tiannuo Wang ◽  
Dongnian Li ◽  
Yang Guo ◽  
...  
Keyword(s):  
Action Recognition ◽  
Three Dimensional ◽  
Experimental Results ◽  
Learning Technology ◽  
Assembly Process ◽  
Convolutional Network ◽  
Data Set ◽  
Batch Normalization ◽  
3D Cnn ◽  
Assembly Error

Monitoring the assembly process is a challenge in the manual assembly of mass customization production, in which the operator needs to change the assembly process according to different products. If an assembly error is not immediately detected during the assembly process of a product, it may lead to errors and loss of time and money in the subsequent assembly process, and will affect product quality. To monitor assembly process, this paper explored two methods: recognizing assembly action and recognizing parts from complicated assembled products. In assembly action recognition, an improved three-dimensional convolutional neural network (3D CNN) model with batch normalization is proposed to detect a missing assembly action. In parts recognition, a fully convolutional network (FCN) is employed to segment, recognize different parts from complicated assembled products to check the assembly sequence for missing or misaligned parts. An assembly actions data set and an assembly segmentation data set are created. The experimental results of assembly action recognition show that the 3D CNN model with batch normalization reduces computational complexity, improves training speed and speeds up the convergence of the model, while maintaining accuracy. Experimental results of FCN show that FCN-2S provides a higher pixel recognition accuracy than other FCNs.

scholarly journals Adversarial Cross-Domain Action Recognition with Co-Attention

2020 ◽  
Vol 34 (07) ◽  
pp. 11815-11822 ◽  
Author(s):  
Boxiao Pan ◽  
Zhangjie Cao ◽  
Ehsan Adeli ◽  
Juan Carlos Niebles
Keyword(s):  
Supervised Learning ◽  
Action Recognition ◽  
Image Recognition ◽  
Attention Mechanism ◽  
Experimental Results ◽  
Single Domain ◽  
Attention Network ◽  
Cross Domain ◽  
The Cross ◽  
Temporal Misalignment

Action recognition has been a widely studied topic with a heavy focus on supervised learning involving sufficient labeled videos. However, the problem of cross-domain action recognition, where training and testing videos are drawn from different underlying distributions, remains largely under-explored. Previous methods directly employ techniques for cross-domain image recognition, which tend to suffer from the severe temporal misalignment problem. This paper proposes a Temporal Co-attention Network (TCoN), which matches the distributions of temporally aligned action features between source and target domains using a novel cross-domain co-attention mechanism. Experimental results on three cross-domain action recognition datasets demonstrate that TCoN improves both previous single-domain and cross-domain methods significantly under the cross-domain setting.

scholarly journals I Know the Relationships: Zero-Shot Action Recognition via Two-Stream Graph Convolutional Networks and Knowledge Graphs

2019 ◽  
Vol 33 ◽  
pp. 8303-8311 ◽  
Author(s):  
Junyu Gao ◽  
Tianzhu Zhang ◽  
Changsheng Xu
Keyword(s):  
Action Recognition ◽  
Visual Cues ◽  
Feature Space ◽  
Experimental Results ◽  
Knowledge Graph ◽  
Convolutional Network ◽  
Convolutional Networks ◽  
Semantic Embedding ◽  
Structured Knowledge ◽  
Knowledge Graphs

Recently, with the ever-growing action categories, zero-shot action recognition (ZSAR) has been achieved by automatically mining the underlying concepts (e.g., actions, attributes) in videos. However, most existing methods only exploit the visual cues of these concepts but ignore external knowledge information for modeling explicit relationships between them. In fact, humans have remarkable ability to transfer knowledge learned from familiar classes to recognize unfamiliar classes. To narrow the knowledge gap between existing methods and humans, we propose an end-to-end ZSAR framework based on a structured knowledge graph, which can jointly model the relationships between action-attribute, action-action, and attribute-attribute. To effectively leverage the knowledge graph, we design a novel Two-Stream Graph Convolutional Network (TS-GCN) consisting of a classifier branch and an instance branch. Specifically, the classifier branch takes the semantic-embedding vectors of all the concepts as input, then generates the classifiers for action categories. The instance branch maps the attribute embeddings and scores of each video instance into an attribute-feature space. Finally, the generated classifiers are evaluated on the attribute features of each video, and a classification loss is adopted for optimizing the whole network. In addition, a self-attention module is utilized to model the temporal information of videos. Extensive experimental results on three realistic action benchmarks Olympic Sports, HMDB51 and UCF101 demonstrate the favorable performance of our proposed framework.

Sign in / Sign up

Export Citation Format

Share Document