Poselet Key-Framing: A Model for Human Activity Recognition

By Michalis Raptis, Leonid Sigal. 2013

244 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Biomedical Engineering, Computer Vision and Pattern Recognition

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 Biomedical Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Unsupervised Learning from Narrated Instruction Videos (2015, background): “Methods in [22, 25] learn in an unsupervised manner the temporal structure of actions from video but do not discover textual expressions for actions as we do in this work.”
An End-to-End Spatio-Temporal Attention Model for Human Action Recognition from Skeleton Data (2016, background): “The learning of key frames can also be cast in a max-margin discriminative framework by treating them as latent variables (Raptis and Sigal 2013).”
A Discriminative Model with Multiple Temporal Scales for Action Prediction (2014, unknown): “The MTSSVM is compared with DBoW and IBoW in [14], the MMED [4], the MSSC and the SC in [1], and the method in [12].”
Bilinear heterogeneous information machine for RGB-D action recognition (2015, methodology): “Related Work Previous action recognition approaches mainly focus on RGB action videos [9, 15, 17, 6].”
ActionVLAD: Learning Spatio-Temporal Aggregation for Action Classification (2017, background): “Other works in event detection and action classification rely on pooling hand crafted features over segments of videos [17, 30, 32, 37, 48].”
Activity Auto-Completion: Predicting Human Activities from Partial Videos (2015, background): “In recent years, predicting human activities from partially observed videos has been an active research topic [18, 8, 17, 13, 14].”
The THUMOS challenge on action recognition for videos "in the wild" (2016, unknown): “…been a few approaches developed for classiﬁcation ( Bojanowski et al., 2014; Duchenne et al., 2009; Karpathy et al., 2014; Niebles et al., 2010; Raptis and Sigal, 2013; Tang et al., 2012 ) and localization ( Hoai et al., 2011; Ke et al., 20 05; 20 07; Pirsiavash and Ramanan, 2014; Tian et al.,…”
Video Action Detection with Relational Dynamic-Poselets (2014, methodology): “The concept of “action part” appeared in several previous works, either implicitly or explicitly [12,13,27,22,26].”
Learning Activity Progression in LSTMs for Activity Detection and Early Detection (2016, methodology): “, by using dynamic bag-of-words of space-time features [19], by modeling actions as a sparse sequence of key-frames [17], or by using compositional kernels to hierarchically capture relationships between partial observations [9].”
Action Recognition Using Rate-Invariant Analysis of Skeletal Shape Trajectories (2016, methodology): “Some of these datasets are also used in the experimental studies described later.”
Predicting the Where and What of Actors and Actions through Online Action Localization (2016, background): “Raptis and Sigal [24] recognize and detect interactions from videos by modeling poselets as latent variables in a structured SVM formulation.”
Video-Based Human Behavior Understanding: A Survey (2013, background): “Raptis and Segal [110] model an action as a very sparse sequence of temporally local discriminative keyframes.”
Dyadic Interaction Detection from Pose and Flow (2014, methodology): “In contrast, we consider videos and take advantage of the movement information by including histograms of oriented flow (HOF) information into the poselet representation, as in [9].”
Action Recognition by Hierarchical Mid-Level Action Elements (2015, methodology): “A large body of work focuses on associating the entire video clip with a single class label from a pre-defined set of action categories (e.g., “take food from fridge” versus “cook food”) (Fig.”
Human Action Recognition With Video Data: Research and Evaluation Challenges (2014, background): “Local patches, thus, generated are characterized by motion patterns [60], [107], optical flow [73], [123], etc.”
Deep Sequential Context Networks for Action Prediction (2017, background): “Existing approaches can be roughly categorized into low-level feature-based approaches [15, 33, 30, 22, 31, 20] and midlevel feature-based approaches [12, 26, 34, 36, 18].”
A Hierarchical Representation for Future Action Prediction (2014, methodology): “…state-of-the-art in recognition at an early stage when only the ﬁrst half of the video is available, and outperforms the current best result [16] For action recognition, our number is slightly lower than state of the art [30], we think it is for two reasons: 1) our method is designed for…”