A survey on still image based human action recognition

By Guo-qing Guo, Alice Lai. 2014

257 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Human-Computer Interaction

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Human-Computer Interaction

Citations in the corpus

RGB-D-based action recognition datasets: A survey (2016, methodology): “Prior to the release of Microsoft Kinect , research has mainly focused on learning and recognizing actions from conventional twodimensional (2D) video [1, 2, 3, 4].”
RGB-D-based Human Motion Recognition with Deep Learning: A Survey (2017, background): “Early research on human motion recognition was dominated by the analysis of still images or videos (Aggarwal and Cai, 1999; Guo and Lai, 2014; Poppe, 2010; Turaga et al., 2008; Wang et al., 2003; Zhu et al., 2016a).”
Neural Motifs: Scene Graph Parsing with Global Context (2017, background): “Compared to object detection [33, 31] , object interactions [43, 3] and activity recognition [13], scene graph parsing poses unique challenges since it requires reasoning about the complex dependencies across all of these components.”
Human Action Recognition From Various Data Modalities: A Review (2020, methodology): “Some survey papers focused on discussing methods using the RGB modality for HAR [5], [30], [31], [32], [33], [34].”
HACS: Human Action Clips and Segments Dataset for Recognition and Temporal Localization (2017, background): “Prior work [37, 32, 21] on exploiting still images for action recognition has shown that still-image classifiers can predict actions in video reasonably well, despite their inability to model motion.”
3D skeleton-based human action classification: A survey (2016, background): “Therefore, there is a vast literature, including interesting surveys [13], [14], [15], [16], [17], [18], [19], on human body pose estimation [20], [21], [22], [23], [24], and activity/action classification [25], [26], [27], [28], [29], [30].”
Describing Common Human Visual Actions in Images (2015, background): “Exploring actions in still images [9] is very valuable given the prevalence and convenience of still pictures.”
Situation Recognition: Visual Semantic Role Labeling for Image Understanding (2016, background): “Activity recognition in still images has been widely studied [21], and it is generally accepted that objects and scenes are important for recognition [31].”
A Comprehensive Survey of Vision-Based Human Action Recognition Methods (2019, methodology): “Because objects and poses provide rich action features for interaction recognition, many methods are based on image data [25].”
Data Augmentation in Classification and Segmentation: A Survey and New Strategies (2023, background): “Motion blur may be used to augment data so as to improve models’ resistance to blurring in action recognition [48,49].”
Semantic Image Retrieval via Active Grounding of Visual Situations (2017, result): “Our situation-retrieval task shares motivation but contrasts with the well-known tasks of “event recognition” or ”action recognition” in still images (e.g., [15]–[17]).”
HICO: A Benchmark for Recognizing Human-Object Interactions in Images (2015, background): “In a literature survey, Guo and Lai [8] reported that the top-used dataset for still image based action recognition (including HOIs) between 2006 and 2013 is Pascal VOC 2010 [6], which contains only 9 categories.”
A Review of Human Activity Recognition Methods (2015, background): “Guo and Lai (2014) summarized all the methods for human activity recognition from still images and categorized them into two big categories according to the level of abstraction and the type of features each method uses.”
Fuzzy human motion analysis: A review (2014, background): “[38] G.”
Learning semantic relationships for better action retrieval in images (2015, background): “[14], most existing action datasets such as the PASCAL actions [10], as well as the Stanford40 [43] are relatively small, with a maximum of 40 actions.”
ActionSpotter: Deep Reinforcement Learning Framework for Temporal Action Spotting in Videos (2020, background): “This issue also prohibits reliance on single-image action classification [12], [13] for such task, as this could lead to discontinuous segments and subsequent overestimation of action occurrences.”