首先,以後可以工作可以考慮一下基於圖卷及的行爲識別今年很多,且在數據集上性能領先。
如下圖所示:
應用場景:
- 【2019】Skeleton-based Action Recognition of People Handling Objects 【論文】
- recognizing object-related human actions
- 偏向於應用場景
- 通過構建skeletion-graph 利用了圖卷積
CNN-Based:
- 【2019】SkeleMotion: A New Representation of Skeleton Joint Sequences Based on Motion Information for 3D Action Recognition【論文】 【代碼】
- Skeleton -> Image
- 偏向於方法
- 3D 行爲識別
- CNN方法
- NTU RGB+D 120 dataset
- skeleton image representation 重要!!!對骨架序列進行編碼,進一步表示,高效表示,skeleton->Image,有點像師兄的思路!
2. 【2019】Three-Stream Convolutional Neural Network With Multi-Task and Ensemble Learning for 3D Action Recognition【論文】
- 數據集:NTU RGBD
- 3D 行爲識別
- 思路:三個stage
- 偏向於方法,網絡結構上的創新!
- 多任務學習
此外他還有一篇文章是考慮了頻域信息進去的,結構差不多
如下,主要講當前的方法都是異步學習語義信息等的,而且是在不同的層,這裏提出了residual frequency domin的attention block!也是網絡結構的上的改進!(多種模態信息的考慮,時-空-頻率)
- Skeleton based action recognition using translation-scale invariant image mapping and multi-scale deep cnn 【論文】
- 對骨架信息轉換成特殊形式的Image
- 多尺度的CNN
- 數據集: NTU RGB+D, UTD-MHAD, MSRC-12, and G3D
- 【2019】Making the Invisible Visible: Action Recognition Through Walls and Occlusions 【論文】
- 對黑暗中或者遮擋住的人體也能構建成骨架,然後識別動作,骨架信息作爲中間信息!
- CNN
- 加入了Attention Module用來獲取時空信息
- 【2018】Co-occurrence Feature Learning from Skeleton Data for Action Recognition and Detection with Hierarchical Aggregation
【論文】
- End- to-end
- CNN
- 數據集:NTU RGB+D, SBU Kinect Interaction and PKU-MMD.
- 首先每個點的信息被單獨的進行了編碼,其次將其assembled into到同時包含時空 信息的高層語音的representation形式,
- 【2019ICME】Learning Shape-Motion Representations from Geometric Algebra Spatio-Temporal Model for Skeleton-Based Action Recognition 【論文】
- 也是對骨架信息的編碼,更高效的表示,然後送到CNN
- 利用了集合代數Geometric Algebra
- multi-stream CNN model
- 數據集:NTU RGB+D and Northwestern-UCLA datasets
- 【2017】3D CNNs on Distance Matrices for Human Action Recognition 【論文】
- 3DCNN+DM(歐式距離矩陣)來獲取的良好的空間幾何結構信息
- 數據集:差不多還是有NTU
- 比之前相近的LSTM-based的方法提升了10個點
- 也可以算是一種對於骨架信息的表示吧
- 【2017】Skeleton-based Action Recognition with Convolutional Neural Networks 【論文】
- CNN+Motion+Trans,還是將骨架序列當做圖片去處理
- action classification and detection
- 骨架序列直接送進CNN之後通過骨架transformer模塊自動選擇中藥的骨架點
- 數據集NTU PKUMMD
- 【PR2017】Enhanced skeleton visualization for view invariant human action recognition-師兄的論文! 【論文】
- Synthesized CNN
- 也是對骨架點進行轉換!之後CNN
- 【2017】A New Representation of Skeleton Sequences for 3D Action Recognition 【論文】
- CNN
- 也是一種對骨架序列重新進行加工的形式使其更加具有表現性!
- A skeleton sequence of any length is transformed into three clips each consisting of several gray images. The generated clips are then fed to a deep CNN model to extract CNN features which are used in a MTLN for action recognition.
- 【2018】A Fine-to-Coarse Convolutional Neural Network for 3D Human Action Recognition 【論文】
- 對骨架序列進行分割成小片段,用來學習其中的聯繫
- 時空信息通過 fine-to-coarse (F2C) CNN architecture 實現
- 數據集:NTU RGB+D and SBU Kinect Interaction dataset.
- 也是對骨架信息動手腳
- 【2017】Interpretable 3D Human Action Analysis with Temporal Convolutional Networks
- CNN
- a way to explicitly learn readily interpretable spatio-temporal representations for 3D human action recognition.
- 【2017】SkeletonNet: Mining Deep Part Features for 3D Action Recognition 【論文】
- 也是一種表現實行,相比於骨架信息本身,處理後的包含了旋轉,評議,和尺度等因素
- 由兩部分構成,第一個用於提取特徵,另一個用來將特徵轉化爲更加具有區分性且緊湊的表現形式
- Two-Stream 3D Convolutional Neural Network for Skeleton-Based Action Recognition-塗塗師姐的論文
- 雙流3DCNN
- 也是一種表現形式,對骨架點編碼到利羣上的點!
- 【2019】Skeleton Image Representation for 3D Action Recognition based on Tree Structure and Reference Joints 【論文】 【代碼】
- a novel skeleton image representation to be used as input to CNNs.
- 數據集:NUTRGBD
- 也是對骨架信息進行處理然後送進CNN
- [2014] Skeletal Quads: Human Action Recognition Using Joint Quadruples 【論文】
- 對骨架信息進行編碼!
- 沒準對之後的工作又用!
- Ensemble One-dimensional Convolution Neural Networks for Skeleton-based Action Recognition
網絡結構的改變 - Hard Sample Mining and Learning for Skeleton-Based Human Action Recognition and Identification 【論文】
- 行爲識別+人的識別identification(應用!)
- 特點:輕量級,快!
混合方法;
- 【2019】Semantics-Guided Neural Networks for Efficient Skeleton-Based Human Action Recognition [【論文】] (https://arxiv.org/pdf/1904.01189v1.pdf)
- e NTU, SYSU, and NUCLA datasets.
- GCN+ CNN
- 也是獲取多種模態信息,如圖
- 【ECCV2018】Skeleton-Based Action Recognition with Spatial Reasoning and Temporal Stack Learning 【論文】 【代碼】
- 圖卷積+LSTM相關,前者獲取高層空間結構信息,後者獲取時域動態信息
- propose a novel model with spatial reasoning and temporal stack learning (SR-TSL) for skeleton based action recognition
RNN-based
1.【2018】 Memory Attention Networks for Skeleton-based Action Recognition 【論文】 【代碼】
- 特點:In this work, we propose a temporal-then-spatial recalibration scheme to alleviate such complex variations, resulting in an end-to-end Memory Attention Networks (MANs) which consist of a Temporal Attention Recalibration Module (TARM) and a Spatio-Temporal Convolution Module (STCM).
- 數據集:NTU RGB+D, HDM05, SYSU-3D and UT-Kinec