PURLS

Introduction

This repository is prepared to provide the code resource for the paper:

Part-aware Unified Representation of Language and Skeleton for Zero-shot Action Recognition by Anqi Zhu, Qiuhong Ke, Mingming Gong, James Bailey.

<div align="center"> <img src="imgs/model_arc.png"> </div>

Latest Updates:

• Uploaded the main model architecture and its relevant package functions. Please visit model/purls.py. (19/06/2024)
• Released pre-print version on arXiv. Available on 21/06/2024. (19/06/2024)

TODOs:

• docs for
  • Prerequisites
  • Demo
  • Data Preparation
  • Testing Pre-trained Models
  • Training
  • Citation
• codes fo\r
  • basic organization and transplantation from implemented codes
  • pre-trained model data
  • data preprocess

Prerequisites

<!-- - Python3 (>3.5) - [PyTorch](http://pytorch.org/) - [Openpose](https://github.com/CMU-Perceptual-Computing-Lab/openpose) **with** [Python API](https://github.com/CMU-Perceptual-Computing-Lab/openpose/blob/master/doc/installation.md#python-api). (Optional: for demo only) - Other Python libraries can be installed by `pip install -r requirements.txt` --> <!-- - FFmpeg (Optional: for demo only), which can be installed by `sudo apt-get install ffmpeg` -->

Installation

<!-- ``` shell git clone https://github.com/yysijie/st-gcn.git; cd st-gcn cd torchlight; python setup.py install; cd .. ``` -->

Get pretrained models

<!-- We provided the pretrained model weithts of our **ST-GCN**. The model weights can be downloaded by running the script ``` bash tools/get_models.sh ``` <!-- The downloaded models will be stored under ```./models```. You can also obtain models from [GoogleDrive](https://drive.google.com/drive/folders/1IYKoSrjeI3yYJ9bO0_z_eDo92i7ob_aF) or [BaiduYun](https://pan.baidu.com/s/1dwKG2TLvG-R1qeIiE4MjeA#list/path=%2FShare%2FAAAI18%2Fst-gcn%2Fmodels&parentPath=%2FShare), and manually put them into ```./models```. -->

Demo

<!-- To visualize how ST-GCN exploit local correlation and local pattern, we compute the feature vector magnitude of each node in the final spatial temporal graph, and overlay them on the original video. **Openpose** should be ready for extracting human skeletons from videos. The skeleton based action recognition results is also shwon thereon. --> <!-- You can use the following commands to run the demo. ```shell # with offline pose estimation python main.py demo_offline [--video ${PATH_TO_VIDEO}] [--openpose ${PATH_TO_OPENPOSE}] # with realtime pose estimation python main.py demo [--video ${PATH_TO_VIDEO}] [--openpose ${PATH_TO_OPENPOSE}] --> <!-- ``` Optional arguments: - `PATH_TO_OPENPOSE`: It is required if the Openpose Python API is not in `PYTHONPATH`. - `PATH_TO_VIDEO`: Filename of the input video. <!-- The realtime demo also support to load video streams from camera source by ``` python main.py demo --video camera ``` --> <!-- Openpose Python API is required in the above demos. -->

Data Preparation

<!--We experimented on two skeleton-based action recognition datasts: **Kinetics-skeleton** and **NTU RGB+D**. Before training and testing, for convenience of fast data loading, the datasets should be converted to proper file structure. You can download the pre-processed data from [GoogleDrive](https://drive.google.com/open?id=103NOL9YYZSW1hLoWmYnv5Fs8mK-Ij7qb) and extract files with ``` cd st-gcn unzip <path to st-gcn-processed-data.zip> ``` Otherwise, for processing raw data by yourself, please refer to below guidances. #### Kinetics-skeleton [Kinetics](https://deepmind.com/research/open-source/open-source-datasets/kinetics/) is a video-based dataset for action recognition which only provide raw video clips without skeleton data. Kinetics dataset include To obatin the joint locations, we first resized all videos to the resolution of 340x256 and converted the frame rate to 30 fps. Then, we extracted skeletons from each frame in Kinetics by [Openpose](https://github.com/CMU-Perceptual-Computing-Lab/openpose). The extracted skeleton data we called **Kinetics-skeleton**(7.5GB) can be directly downloaded from [GoogleDrive](https://drive.google.com/open?id=1SPQ6FmFsjGg3f59uCWfdUWI-5HJM_YhZ) or [BaiduYun](https://pan.baidu.com/s/1dwKG2TLvG-R1qeIiE4MjeA#list/path=%2FShare%2FAAAI18%2Fkinetics-skeleton&parentPath=%2FShare). After uncompressing, rebuild the database by this command: ``` python tools/kinetics_gendata.py --data_path <path to kinetics-skeleton> ``` #### NTU RGB+D NTU RGB+D can be downloaded from [their website](http://rose1.ntu.edu.sg/datasets/actionrecognition.asp). Only the **3D skeletons**(5.8GB) modality is required in our experiments. After that, this command should be used to build the database for training or evaluation: ``` python tools/ntu_gendata.py --data_path <path to nturgbd+d_skeletons> ``` where the ```<path to nturgbd+d_skeletons>``` points to the 3D skeletons modality of NTU RGB+D dataset you download.-->

Testing Pretrained Models

<!-- ### Evaluation Once datasets ready, we can start the evaluation. To evaluate ST-GCN model pretrained on **Kinetcis-skeleton**, run ``` python main.py recognition -c config/st_gcn/kinetics-skeleton/test.yaml ``` For **cross-view** evaluation in **NTU RGB+D**, run ``` python main.py recognition -c config/st_gcn/ntu-xview/test.yaml ``` For **cross-subject** evaluation in **NTU RGB+D**, run ``` python main.py recognition -c config/st_gcn/ntu-xsub/test.yaml ``` <!-- Similary, the configuration file for testing baseline models can be found under the ```./config/baseline```. To speed up evaluation by multi-gpu inference or modify batch size for reducing the memory cost, set ```--test_batch_size``` and ```--device``` like: ``` python main.py recognition -c <config file> --test_batch_size <batch size> --device <gpu0> <gpu1> ... ``` ### Results The expected **Top-1** **accuracy** of provided models are shown here: | Model| Kinetics-<br>skeleton (%)|NTU RGB+D <br> Cross View (%) |NTU RGB+D <br> Cross Subject (%) | | :------| :------: | :------: | :------: | |Baseline[1]| 20.3 | 83.1 | 74.3 | |**ST-GCN** (Ours)| **31.6**| **88.8** | **81.6** | [1] Kim, T. S., and Reiter, A. 2017. Interpretable 3d human action analysis with temporal convolutional networks. In BNMW CVPRW. -->

Training

main.py also supports training a new model with customized configs. The script accepts the following parameters:

| Argument | Possible Values | Description | --- | --- | --- | ntu | 60; 120 | Which NTU dataset to use | ss | 5; 12 (For NTU-60); 24 (For NTU-120) | Which split to use | st | r (for random) | Split type | phase | train; val | train(required for zsl), (once with train and once with val for gzsl) | ve | shift; msg3d | Select the Visual Embedding Model | le | w2v; bert | Select the Language Embedding Model | num_cycles | Integer | Number of cycles(Train for 10 cycles) | num_epoch_per_cycle | Integer | Number of epochs per cycle 1700 for 5 random and 1900 for others| latent_size | Integer | Size of the skeleton latent dimension (100 for ntu-60 and 200 for ntu-120)| load_epoch | Integer | The epoch to be loaded | load_classifier | | Set if the pre-trained classifier is to be loaded | dataset |- | Path to the generated visual features | wdir | - | Path to the directory to store the weights in | mode | train;eval | train for training synse, eval to eval using a pretrained model | gpu | - | which gpu device number to train on |

For example, if you want to train PURLS for zsl under a experiment split of 55/5 split on NTU 60, you can use the following command: <code> python main.py -c configs/adaptive_purls_5r_clip_gb.yml </code>

<!--To train a new ST-GCN model, run ``` python main.py recognition -c config/st_gcn/<dataset>/train.yaml [--work_dir <work folder>] ``` where the ```<dataset>``` must be ```ntu-xsub```, ```ntu-xview``` or ```kinetics-skeleton```, depending on the dataset you want to use. The training results, including **model weights**, configurations and logging files, will be saved under the ```./work_dir``` by default or ```<work folder>``` if you appoint it. You can modify the training parameters such as ```work_dir```, ```batch_size```, ```step```, ```base_lr``` and ```device``` in the command line or configuration files. The order of priority is: command line > config file > default parameter. For more information, use ```main.py -h```. Finally, custom model evaluation can be achieved by this command as we mentioned above: ``` python main.py recognition -c config/st_gcn/<dataset>/test.yaml --weights <path to model weights> ``` -->

Citation

<!-- Please cite the following paper if you use this repository in your reseach. --> <!-- ``` @inproceedings{stgcn2018aaai, title = {Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition}, author = {Sijie Yan and Yuanjun Xiong and Dahua Lin}, booktitle = {AAAI}, year = {2018}, } ``` -->

Contact

For any question, feel free to create a new issue or contact.

Qiuhone Ke     : qiuhong.ke@monash.edu
Anqi Zhu : azzh1@student.unimelb.edu.au