Skip to content

Latest commit

 

History

History
88 lines (65 loc) · 4.7 KB

readme.md

File metadata and controls

88 lines (65 loc) · 4.7 KB
Raw

PointLSTM

This repo holds the codes of paper: An Efficient PointLSTM for Point Clouds Based Gesture Recognition (CVPR 2020). [supp][video]

Abstract

Point clouds contain rich spatial information, which provides complementary cues for gesture recognition. In this paper, we formulate gesture recognition as an irregular sequence recognition problem and aim to capture long-term spatial correlations across point cloud sequences. A novel and effective PointLSTM is proposed to propagate information from past to future while preserving the spatial structure. The proposed PointLSTM combines state information from neighboring points in the past with current features to update the current states by a weight-shared LSTM layer. This method can be integrated into many other sequence learning approaches. In the task of gesture recognition, the proposed PointLSTM achieves state-of-the-art results on two challenging datasets (NVGesture and SHREC’17) and outperforms previous skeleton-based methods. To show its advantages in generalization, we evaluate our method on MSR Action3D dataset, and it produces competitive results with previous skeleton-based methods.

tease_img

Prerequisites

These code is implemented in Pytorch (>1.0). Thus please install Pytorch first.

Usage

Get the code

Clone this repo with git, please use:

git clone https://github.com/Blueprintf/pointlstm_gesture_recognition_pytorch.git

Data Preparation

SHREC'17

  • Download the SHREC'17 dataset [Visualization] and put HandGestureDataset_SHREC2017 directory to ./dataset/SHREC2017, It is suggested to make a soft link toward downloaded dataset.
  • Generate point cloud sequences from depth video, and save the processed point clouds in ./dataset/Processed_SHREC2017. Each video generate 32*256 points, and the generated point clouds occupy about 2.5G.
cd dataset
python shrec17_process.py

NvGesture

  • Download the NvGesture dataset [visualization] and extract the NvGesture directory to ./dataset/Nvidia, it is suggested to make a soft link toward downloaded dataset.
  • Generate point cloud sequences from depth video, and save the processed point clouds in ./dataset/Nvidia/Processed. Each video generate 32* 512 points, and the generated point clouds occupy about 11G.
cd dataset
python nvidia_dataset_split.py
python nvidia_process.py

Training

Training of the PointLSTM-middle with k=16 on SHREC'17:

cd experiments
python main.py --phase=train --work-dir=PATH_TO_SAVE_RESULTS --device=0

We also provided trained model at here Google Drive or Baidu Yun [passwd: trhi].

Inference

cd experiments
python main.py --phase=test --work-dir=PATH_TO_SAVE_RESULTS --device=0 --weights=PATH_TO_WEIGHTS

Citation

Please cite the following paper if you feel PointLSTM useful to your research.

@inproceedings{min_CVPR2020_PointLSTM,
  title={An Efficient PointLSTM for Point Clouds Based Gesture Recognition},
  author={Min, Yuecong and Zhang, Yanxiao and Chai, Xiujuan and Chen, Xilin},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={5761--5770},
  year={2020}
}

Relevant paper: FlickerNet: Adaptive 3D Gesture Recognition from Sparse Point Clouds. [paper]

@inproceedings{min2019flickernet,
  title={FlickerNet: Adaptive 3D Gesture Recognition from Sparse Point Clouds.},
  author={Min, Yuecong and Chai, Xiujuan and Zhao, Lei and Chen, Xilin},
  booktitle={BMVC},
  pages={105},
  year={2019}
}

Reference

[1] Yan, Sijie, Yuanjun Xiong, and Dahua Lin. "Spatial temporal graph convolutional networks for skeleton-based action recognition." Thirty-second AAAI conference on artificial intelligence. 2018. [pdf] [code]