Multi-region Two-Stream R-CNN for Action Detection

By Xiaojiang Peng, Cordelia Schmid. 2016

394 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Computer Vision and Pattern Recognition

Citations in the corpus

RGB-D-based Human Motion Recognition with Deep Learning: A Survey (2017, background): “Peng and Schimd [96] generated rich proposals from both RGB and optical flow data by using region proposal networks for frame-level action detection, and stacked optical flows to enhance the discriminative power of motion R-CNN.”
Chained Multi-stream Networks Exploiting Pose, Motion, and Appearance for Action Classification and Detection (2017, methodology): “Following recent works on action detection [10, 48, 31], we report video-AP.”
Tube Convolutional Neural Network (T-CNN) for Action Detection in Videos (2017, methodology): “[19] in the plot, since they do not provide the ROC and AUC curves.”
Evaluate the Malignancy of Pulmonary Nodules Using the 3-D Deep Leaky Noisy-OR Network (2017, methodology): “Therefore, some methods use 2-D region proposal networks (RPNs) to extract proposals in individual 2-D images then combine them to generate 3-D proposals [8], [9].”
TORNADO: A Spatio-Temporal Convolutional Regression Network for Video Action Proposal (2017, methodology): “Saha et al.[32] and Peng et al.[25] extends the approach in [9] by using Fast-RCNN which first detect class-specific action regions, and then seam them to be class-specific action tubes using dynamic programming.”
Rethinking Spatiotemporal Feature Learning: Speed-Accuracy Trade-offs in Video Classification (2017, methodology): “Similar to the framework proposed in [49], we use the Faster-RCNN [50] object detection algorithm to jointly perform person localization and action recognition.”
Temporal Action Detection with Structured Segment Networks (2017, background): “Note there are also works on spatial-temporal detection [13, 54, 25, 50, 31] and temporal video segmentation [15], which are beyond the scope of this paper.”
Attend and Interact: Higher-Order Object Interactions for Video Understanding (2017, background): “Another branch of work uses Region Proposal Networks (RPNs) to jointly train action detection models [15, 25, 36].”
Action Tubelet Detector for Spatio-Temporal Action Localization (2017, methodology): “Following standard practice for action localization [23, 27, 36], we use a two-stream detector.”
Action Search: Spotting Actions in Videos and Its Application to Temporal Action Localization (2017, background): “A large number of works have successfully tackled the task of action recognition [8,9,28,38,42] and spatio-temporal localization [10,18,27,30,33,39].”
Trespassing the Boundaries: Labeling Temporal Bounds for Object Interactions in Egocentric Video (2017, background): “Recently, Peng et al. [20] assessed how frame-level clas-siﬁcations using multi-region two-stream CNN are pooled to achieve video-level recognition results.”
Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset (2017, methodology): “Recent works have proposed imaginative methods for determining the spatial and temporal extent (detection) of actors within the two-stream architectures, by incorporating linked object detections in time [22, 24].”
A flexible model for training action localization with varying levels of supervision (2018, methodology): “Currently, the dominant strategy [11, 31, 36, 37, 45] is to obtain perframe action detections and then to link them into continuous spatio-temporal tracks.”
Human Action Localization with Sparse Spatial Supervision (2017, methodology): “The drop in performance compared to for example [10] can be explained by their additional multi-region and multi-flow description.”
Online Real-Time Multiple Spatiotemporal Action Localisation and Prediction (2016, methodology): “Contributions ➢ Unlike previous methods [1,2,3], we construct multiple action tubes simultaneously and incrementally, getting rid of recursive calls to dynamic programming to generate multiple tubes.”
Rethinking Spatiotemporal Feature Learning For Video Understanding (2017, methodology): “Similar to the framework proposed by Peng and Schmid [31], we use the Faster-RCNN [36] object detection algorithm to jointly perform actor localization and action recognition.”
Pattern Recognition (2015, neutral): “For [28] , we report the experimental results with and without the multi-region method.”
PoTion: Pose MoTion Representation for Action Recognition (2018, neutral): “5%) than the classification performance reported by action localization approaches [14, 29] and 1.”