Action snippets: How many frames does human action recognition require?

By Konrad Schindler, Luc Van Gool. 2008

620 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Artificial Intelligence

In Computer Vision and Pattern Recognition

In Human-Computer Interaction

Citations in the corpus

Lifelong learning of human actions with deep neural network self-organization (2017, methodology): “For compatibility with Schindler and Van Gool (2008), we trimmed all sequences to a total of 28 frames,which is the length of the shortest sequence, and evaluated our approach by performing leave-one-out cross-validation, i.e., 8 subjects were used for training and the remaining one for testing.”
Video Representation Learning Using Discriminative Pooling (2018, background): “To this end, we observe that not all predictions on the short video snippets are equally informative, yet some of them must be [36].”
Recognizing Actions in 3D Using Action-Snippets and Activated Simplices (2016, methodology): “We incorporate the temporal order by combining consecutive poses together to form an action-snippet ˆ y (Schindler and Van Gool 2008) (ten poses give the best results, see later).”
Trespassing the Boundaries: Labeling Temporal Bounds for Object Interactions in Egocentric Video (2017, background): “Similarly, two works evaluated the optimal segment length for action recognition [25, 36].”
From Actemes to Action: A Strongly-Supervised Representation for Detailed Action Understanding (2013, background): “Here, parts have been proposed that capture a neighborhood of spacetime [13, 7, 29, 22], a spatial keyframe [26] or a temporal segment while considering the entire spatial extent of the scene [21, 9].”
Recognizing actions by shape-motion prototype trees (2009, methodology): “Our approach leave one out 95.77 93.43 Schindler [ 23 ] split 90.73 92.7”
3C-Net: Category Count and Center Loss for Weakly-Supervised Action Localization (2019, background): “Further, unlike object boundary annotations in images, manual annotations of temporal action boundaries are more subjective and prone to large variations [20, 18].”
Real-time Action Recognition by Spatiotemporal Semantic and Structural Forests (2010, background): “as suggested by [22], actions can be recognised from very short sequences called the “snippets”.”
Human Action Recognition by Semilatent Topic Models (2009, neutral): “Schindler and Van Gool [ 30 ] perform action recognition by training SVM classifiers based on local shapes and dense optical flows.”
HMDB: A large video database for human motion recognition (2011, background): “Clips extracted from the same movie usually have similar scenes.”
Temporal Localization of Actions with Actoms (2013, neutral): “For instance, Efros et al .”
Neural representation of action sequences: how far can a simple snippet-matching model take us? (2013, result): “The results are surprisingly good for such a simple model, consistent with findings from computer vision [20].”
A survey on vision-based human action recognition (2010, background): “Schindler and Van Gool [124] extend the work by Jhuang et al.”
Exploring the Trade-off Between Accuracy and Observational Latency in Action Recognition (2012, methodology): “In previous work, such as the Action Snippets proposed by Schindler and Van Gool [28], and the work of Davis and Tyagi [7], the system is tested on sequences where the action is being performed continuously (no transition from a rest position), ensuring that every subset of frames shows the action i”