# 正確安裝STRIVE完整環境 Setting Up STRIVE Environment

This guide provides step-by-step instructions to set up the STRIVE environment, ensuring compatibility and successful installation. STRIVE is a codebase for the CVPR 2022 paper "Generating Useful Accident-Prone Driving Scenarios via a Learned Traffic Prior."

## Prerequisites
- Ubuntu 18.04
- Python 3.6
- CUDA 11.1

### Important Note !!!
Before installing the requirements, you **must** install `opencv-python==4.6.0.66`. Failure to do so may lead to issues with the nuScenes devkit and incorrect OpenCV installation, preventing successful setup.

## Steps to Set Up the Environment

### 1. Create and Activate a Virtual Environment
First, create a virtual environment named `strive_env` and activate it:

python3 -m venv strive_env
source strive_env/bin/activate

### 2. Update Pip
Upgrade pip to ensure you have the latest version:

pip install --upgrade pip

### 3. Install Numpy
Install numpy with the specified version:

pip install numpy==1.19.5

### 4. Install OpenCV
Install the specific version of OpenCV to avoid compatibility issues:

pip install opencv-python==4.6.0.66

### 5. Install Remaining Dependencies
Install the remaining dependencies from the `requirements.txt` file. Note that the exact versions in the requirements file are crucial for compatibility:

pip install -r requirements.txt

## Download and Prepare the nuScenes Dataset
Download the nuScenes dataset (only the metadata is required) and the map expansion. Place them in a directory named `data` following the structure below(this is the correct file structure):

data/nuscenes/trainval
|______ v1.0-trainval
        |___________*.json
|______ maps
        |___________ basemap
        |___________ expansion
        |___________ prediction
        |___________ *.png

## Download Technique !!
Download MiniSet, Map expansion1.3 and or choose between whole data (25gb up) can make the dataset structure complete.

## Training and Testing the Traffic Model

### Training
To train the traffic model from scratch on the car and truck categories, run:

python src/train_traffic.py --config ./configs/train_traffic.cfg

### Testing
Evaluate performance and visualize samples after training:

python src/test_traffic.py --config ./configs/test_traffic.cfg --ckpt path/to/model.pth

## Further Information
For more information and results, check out the [project page](https://nv-tlabs.github.io/STRIVE/).

## Contact
[e-mail]