Action Recognition Based on Sub-action Motion History Image and Static History Image

Zhang, Shichao; Chen, Enqing; Chen, Qi; Liang, Chengwu

doi:10.1051/matecconf/20165602006

Cited by 5 publications

(3 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For S-CNN, the temporal segment windows were set to 16, 32, 64, 128, 256 and 512 frames with 75% overlap. For R-C3D, the frames at every 1/8 of the length of a video were set as the anchors, and the value of K in the algorithm was set to (4,5,6,7,8,9,10,11,12,13,14,15,16).…”

Section: Talmentioning

confidence: 99%

“…Consequently, work on many tasks, including HAR and TAL, has undergone dramatic development. 2D-CNN-based methods have been used to learn discriminative features from pre-processed motion images, such as optical flow (OF) features, 9,10 motion history images (MHIs), 11 static history images (SHIs), 12 motion energy images (MEIs) 13 and other variants. [14][15][16][17][18][19][20][21] These classic motion images can handle temporal information well and compress information from multiple frames into a motion image with a uniform size.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Multi‐scale feature learning and temporal probing strategy for one‐stage temporal action localization

Yao

Yang

Huang

et al. 2021

Int J of Intelligent Sys

View full text Add to dashboard Cite

The aim of temporal action localization (TAL) is to determine the start and end frames of an action in a video. In recent years, TAL has attracted considerable attention because of its increasing applications in video understanding and retrieval. However, precisely estimating the duration of an action in the temporal dimension is still a challenging problem. In this paper, we propose an effective one‐stage TAL method based on a self‐defined motion data structure, called a dense joint motion matrix (DJMM), and a novel temporal detection strategy. Our method provides three main contributions. First, compared with mainstream motion images, DJMMs can preserve more pre‐processed motion features and provides more precise detail representations. Furthermore, DJMMs perfectly solve the temporal information loss problem caused by motion trajectory overlaps within a certain time period. Second, a spatial pyramid pooling (SPP) layer, which is widely used in the object detection and tracking fields, is innovatively incorporated into the proposed method for multi‐scale feature learning. Moreover, the SPP layer enables the backbone convolutional neural network (CNN) to receive DJMMs of any size in the temporal dimension. Third, a large‐scale‐first temporal detection strategy inspired by a well‐developed Chinese text segmentation algorithm is proposed to address long‐duration videos. Our method is evaluated on two benchmark data sets and one self‐collected data set: Florence‐3D, UTKinect‐Action3D and HanYue‐3D. The experimental results show that our method achieves competitive action recognition accuracy and high TAL precision, and its time efficiency and few‐shot learning capabilities enable it to be utilized for real‐time surveillance.

show abstract

Section: Talmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Multi‐scale feature learning and temporal probing strategy for one‐stage temporal action localization

Yao

Yang

Huang

et al. 2021

Int J of Intelligent Sys

View full text Add to dashboard Cite

show abstract

“…Inspired by the paper [7] the depth of the image in the three Cartesian plane projection followed by the extraction of the DMM images of each plane so the body action recognition. In this paper a fast and accurate sitting posture recognition algorithm is proposed based on the human body sitting posture depth image.…”

Section: Feature Extractionmentioning

confidence: 99%