Action Recognition and Detection by Combining Motion and Appearance Features

By Limin Wang, Yu Qiao, Xiaoou Tang. 2014

How researchers see this paper

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

Citations in the corpus

PKU-MMD: A Large Scale Benchmark for Continuous Multi-Modal Human Action Understanding (2017, methodology): “For action detection, existing methods mainly utilize either sliding-window scheme [46, 31], or action proposal approaches [47].”
Efficient Action Detection in Untrimmed Videos via Multi-task Learning (2016, background): “While this pipeline has achieved state-of-the-art performance in recent years [23, 36, 16, 45], we argue that learning and applying the closely related steps independently has several drawbacks.”
CDC: Convolutional-De-Convolutional Networks for Precise Temporal Action Localization in Untrimmed Videos (2017, methodology): “A typical framework used by many state-of-the-art systems [67, 53, 39, 65, 26] is fusing a large set of features and training classifiers that operate on sliding windows or segment proposals.”
Predictive-Corrective Networks for Action Detection (2017, methodology): “Temporal action localization: A common way to extend action recognition models to temporal detection is through the sliding windows paradigm [48, 17, 49, 27, 57].”
A Survey on Deep Learning Based Approaches for Action and Gesture Recognition in Image Sequences (2017, unknown): “2014 THUMOS-14 A ECCV [44] [46] [95] [111] [88]”
SST: Single-Stream Temporal Action Proposals (2017, background): “A number of existing methods approach the problem of temporal action localization [19, 22, 26, 27, 31, 32, 35, 40, 24, 30, 38].”
CUHK & SIAT Submission for THUMOS 15 Action Recognition Challenge (2015, methodology): “In order to perform action recognition in temporal untrimmed videos, we follow our previous method [14] and first temporally divide continuous videos into short clips.”
R-C3D: Region Convolutional 3D Network for Temporal Activity Detection (2017, methodology): “Existing state-of-the-art approaches address this task as detection by classiﬁcation , i.e. classifying temporal segments generated in the form of sliding windows [13, 20, 24, 37] or via an external “pro-posal” generation mechanism [10, 35].”
Oops! Predicting Unintentional Action in Video (2019, background): “Critical to this success are datasets of diverse videos released to facilitate this research [52, 6, 34, 55, 66, 29, 8, 1, 43, 53, 30, 19, 18, 13, 20, 42].”
Temporal Action Localization in Untrimmed Videos via Multi-stage CNNs (2016, methodology): “For example, top performing approaches in THUMOS Challenge 2014 [27, 41, 17, 15] and 2015 [46, 9] both used improved Dense Trajectory (iDT) with Fisher Vector (FV) [40, 25].”
One-Shot Action Localization by Learning Sequence Matching Network (2018, background): “While early attempts relied on handcrafted video features [31, 19, 44, 25], deep neural network based approaches have made significant progress [9, 20, 35, 48, 49, 34] and reported the state-of-the-art performance due to data-driven spatiotemporal feature representations, effecFigure 1.”