Skip to content

Latest commit

 

History

History
108 lines (96 loc) · 3.67 KB

Env_setup.md

File metadata and controls

108 lines (96 loc) · 3.67 KB
Raw

Environment Setup

We can use either Docker or Conda for building environment.

Environment setup with Dockerfile

We recommend to use DockerFile for easy reproduction.

  1. First, check if you have docker and Nvidia-cuda setup.If not, follow the official website to setup docker:
sudo docker run --rm --gpus all nvidia/cuda:11.0-base nvidia-smi
  1. Build Docker file:
docker build -t cctrack3 docker/
  1. Run it with:
docker run --gpus all --shm-size=200g -it -v {DATA_DIR}:/mmdetection/data cctrack3

where DATA_DIR=<path-to-AI-CITY-2022-Track3-Downloaded folder>. We also provide the script to run docker file.

./docker/run_docker.sh

Environment Setup with CONDA

  1. Create a conda virtual environment and activate it.
conda create -n aicity_t3 python=3.7 -y
conda activate aicity_t3
  1. Install PyTorch and torchvision following the official instructions, e.g.,
conda install pytorch torchvision -c pytorch

Note: Make sure that your compilation CUDA version and runtime CUDA version match. You can check the supported CUDA version for precompiled packages on the PyTorch website.

E.g.1 If you have CUDA 10.2 installed under /usr/local/cuda and would like to install PyTorch 1.5, you need to install the prebuilt PyTorch with CUDA 10.2.

conda install pytorch cudatoolkit=10.2 torchvision -c pytorch
  1. Install MMAction2 Install MMAction2 with MIM.
pip install openmim
mim install mmaction2==0.17.0

MIM can automatically install OpenMMLab projects and their requirements.

  1. Install CCAction package
python setup.py develop
  1. Install Actionformer package
cd actionformer/utils
python setup.py install --user
cd ../..

Data Download and extract frame:

We assume that the dataset is downloaded and placed (or have symbolic-link) in the following structure :

├── ccaction/
├── configs/
├── README.md
├── reproduce_scripts/
├── setup.py
├── tools/
├── data/ 
    └──raw_video/ -> <download_folder>/ai-city-2022_track3
        ├── A1/
        ├── A2/
        ├── Distracted_Activity_Class_definition.txt
        ├── README.txt
        └── video_ids.csv

[Optional] For reproducing purpose and quick evaluation, we also provide our Manual Labels for the A2 set in:

reproduce_scripts/A2_local_labels

Please copy the corresponding labels file for each user to the folder A2 similar to A1 folder:

data/raw_video/A2
                ├── user_id_42271
                │   ├── Dashboard_user_id_42271_NoAudio_3.MP4
                │   ├── Dashboard_user_id_42271_NoAudio_4.MP4
                │   ├── Rear_view_user_id_42271_NoAudio_3.MP4
                │   ├── Rear_view_user_id_42271_NoAudio_4.MP4
                │   ├── Right_side_window_user_id_42271_NoAudio_3.MP4
                │   ├── Right_side_window_user_id_42271_NoAudio_4.MP4
                │   └── user_id_42271.csv
                ├── user_id_56306
                │   ├── Dashboard_user_id_56306_NoAudio_2.MP4
                │   ├── Dashboard_user_id_56306_NoAudio_3.MP4
                │   ├── Rear_view_user_id_56306_NoAudio_2.MP4
                │   ├── Rear_view_user_id_56306_NoAudio_3.MP4
                │   ├── Rightside_window_user_id_56306_NoAudio_2.MP4
                │   ├── Rightside_window_user_id_56306_NoAudio_3.MP4
                │   └── user_id_56306.csv
                ...