3D Action Recognition from Novel Viewpoints

By Hossein Rahmani, Ajmal Saeed Mian. 2016

202 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Artificial Intelligence

1. Jun Liu
2. G. Wang
3. Hossein Rahmani
4. Cordelia Schmid
5. Andrew Zisserman
6. Li Fei-Fei
7. Luc Van Gool
8. Abhinav Kumar Gupta

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. G. Wang
3. Jitendra Malik
4. Dacheng Tao
5. Ming-Hsuan Yang
6. Andrew Zisserman
7. Cordelia Schmid
8. Li Fei-Fei

In 3D Vision & Pose

1. Jun Liu
2. G. Wang
3. Hossein Rahmani
4. Amir Shahroudy
5. Michael J. Black
6. Tae-Kyun Kim
7. Pichao Wang
8. Ivan Laptev

Citations in the corpus

• NTU RGB+D: A Large Scale Dataset for 3D Human Activity Analysis (2016, methodology): “To apply ConvNet-based learning to this domain, [30] used synthetically generated data and fitted them to real mocap data.”
• Deep Multimodal Feature Analysis for Action Recognition in RGB+D Videos (2016, methodology): “To apply ConvNetbased learning to this domain, [49] used synthetically generated data and fitted them to real mocap data.”
• Glimpse Clouds: Human Activity Recognition from Unstructured Feature Points (2018, neutral): “2 - - HPM+TM [43] D 91.”
• RGB-D-based Human Motion Recognition with Deep Learning: A Survey (2017, methodology): “Most of available datasets require subjects to perform actions in a visible and restricted view to avoid occlusion, and this results in limited view data collection and less 29 SBU Kinect Interaction [149] Depth+ Skeleton PRNN Side Information 89.20% [130] Skeleton Co-occurrence +LSTM None 90.41% [1”
• Unsupervised Learning of Long-Term Motion Dynamics for Videos (2017, background): “Skeleton-based approaches such as [5, 22, 35, 39, 50] move beyond such sparse grid-like pooling and focus on how to propose good skeletal representations.”
• Sim2real transfer learning for 3D human pose estimation: motion to the rescue (2019, background): “Other works consider pedestrian detection [59–61] and action recognition [62, 63].”
• Configurable 3D Scene Synthesis and 2D Image Rendering with Per-pixel Ground Truth Using Stochastic Grammars (2017, background): “, 119], 2D human pose estimation [93, 96, 105], 3D human pose estimation [18, 27, 39, 104, 109, 111, 126, 139, 151], depth prediction [118], pedestrian detection [49, 81, 94, 127], action recognition [100, 101, 115], semantic segmentation [103], sceneunderstanding[45, 46, 60, 97],aswellasinbenchmark”
• Human Action Recognition From Various Data Modalities: A Review (2020, methodology): “Besides, Skepxels were used to hierarchically capture the micro-temporal relations between the joints in the frames, and Fourier Temporal Pyramids [9]”
• Enhanced skeleton visualization for view invariant human action recognition (2017, unknown): “Depth Hand-crafted HON4D [44] 39.90 SNV [45] 42.80 AOG [39] 53.60 HOPC [40] 80.00 CNN HPM + TM [46] 92.00 Hybrid Hand-crafted AE [47] 76.00 Skeleon Hand-crafted HOJ3D [17] 54.50 LARP [27] 74. method.”
• Spatio-Temporal Dynamics and Semantic Attribute Enriched Visual Encoding for Video Captioning (2019, background): “It is noteworthy that exploiting Fourier Transform in a hierarchical fashion to encode temporal dynamics has also been considered in human action recognition [53, 36].”
• Learning a Deep Model for Human Action Recognition from Novel Viewpoints (2016, background): “Due to the impressive results of such deep learning on handwritten digit recognition [48], image classification [50] and object detection [51], several methods [52], [53], [54], [55], [56] have been recently proposed to learn deep models for video based action recognition.”
• Learning from Synthetic Humans (2017, background): “[31, 37] train CNNs with synthetic depth images.”
• First-Person Hand Action Benchmark with RGB-D Videos and 3D Hand Pose Annotations (2017, methodology): “These methods usually focus on how to extract discriminative features from the depth images using local geometric descriptors [40, 43, 76] sensitive to viewpoint changes and viewinvariant approaches [46, 47].”
• H+O: Unified Egocentric Recognition of 3D Hand-Object Poses and Interactions (2019, background): “e error. Model Input modality Accuracy Two-stream-color [15] Color 61.56 Two-stream-flow [15] Color 69.91 Two-stream-all [15] Color 75.30 Joule-color [24] Color 66.78 HON4D [42] Depth 70.61 Novel View [47] Depth 69.21 Joule-depth [24] Depth 60.17 [17] + Gram Matrix Depth 32.22 [17] + Lie Group Depth ”
• A Survey on Deep Learning Based Approaches for Action and Gesture Recognition in Image Sequences (2017, methodology): “Differently, [85] uses CNNs to obtain canonical human poses for action recognition.”