Fast action proposals for human action detection and search

By Gang Yu, Junsong Yuan. 2015

222 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, 3D Vision & Pose

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 3D Vision & Pose

Citations in the corpus

Object-Proposal Evaluation Protocol is ‘Gameable’ (2015, background): “Then average recall is measured over all the ground truth instances: where I [ · ] is an indicator function for the logical preposition in the argument.”
Chained Multi-stream Networks Exploiting Pose, Motion, and Appearance for Action Classification and Detection (2017, background): “Compared to the recent works [10, 49, 31, 51], our detection framework has two desirable properties: (1) the pose network directly provides a single detection box per person, which causes a large speed-up; (2) the classification takes advantage of three modalities and the chained fusion, which yield”
End-to-End Learning of Action Detection from Frame Glimpses in Videos (2015, background): “It must learn to effectively utilize these observations, or frame glimpses, to reason on the temporal bounds of actions.”
Online Human Action Detection using Joint Classification-Regression Recurrent Neural Networks (2016, methodology): “To localize actions in streaming video sequence, existing detection methods utilize either sliding-window scheme [8,5,9,10], or action proposal approaches [20,21,11].”
Towards Weakly-Supervised Action Localization (2016, background): “In [12], proposals are based on an action-ness measure [24], which requires localized training samples.”
Learning to Track for Spatio-Temporal Action Localization (2015, neutral): “At a threshold δ = 5%, we obtain a mAP of 54% compared to 42% reported by [46].”
TORNADO: A Spatio-Temporal Convolutional Regression Network for Video Action Proposal (2017, background): “Different approaches have been explored including segmentation-and-merging [11, 22, 19], dense trajectories clustering [39], human detection [42], deep learning [16], etc.”
TURN TAP: Temporal Unit Regression Network for Temporal Action Proposals (2017, background): “Number of retrieved proposals (AR-N) [31, 9], Average Recall vs.”
Temporal Context Network for Activity Localization in Videos (2017, methodology): “For spatio-temporal localization of actions, multiple methods use spatio-temporal region proposals [9, 22, 7, 39].”
Object Detection from Video Tubelets with Convolutional Neural Networks (2016, methodology): “The methods that are based on action proposals are related to our work.”
Efficient Action Detection in Untrimmed Videos via Multi-task Learning (2016, background): “Note that, [43] and [34] are designed to generate spatio-temporal proposals.”
SST: Single-Stream Temporal Action Proposals (2017, unknown): “[4, 12, 16, 34, 37, 39].”
Temporal Action Localization in Untrimmed Videos via Multi-stage CNNs (2016, background): “Recently, researchers treat this as a tracking problem [43, 8] by leveraging object detectors [11], especially human detectors [16, 19, 8, 45] to detect regions of interest in each frame and then output sequences of bounding boxes.”
Vision-based human activity recognition: a survey (2020, methodology): “These methods allow the recognition of complex human activities using complex classification models, which make the specification of emotional attributes difficult [144, 228].”
Leaving Some Stones Unturned: Dynamic Feature Prioritization for Activity Detection in Streaming Video (2016, background): ““Action-like” space-time proposals [31]–[34] and efficient search methods [35], [36] can avoid applying classifiers to all possible video subvolumes, but they do not prioritize feature computation.”
CDC: Convolutional-De-Convolutional Networks for Precise Temporal Action Localization in Untrimmed Videos (2017, methodology): “Various methods have been developed, from the perspective of supervoxel merging [20, 54, 55], tracking [69, 42, 62, 52], object detection and linking [28, 14, 75, 42, 62], spatio-temporal segmentation [31, 70], and leveraging still images [21, 58, 22].”
Spatial-Aware Object Embeddings for Zero-Shot Localization and Classification of Actions (2017, methodology): “Action proposal methods include merging supervoxels [18, 44], merging trajectories [3, 35], and detecting actors [63].”
AMTnet: Action-Micro-Tube Regression by End-to-end Trainable Deep Architecture (2017, methodology): “We can observe that our model outperforms [36, 33, 22] by a large margin.”