Risk_identification_tool

Setup

1. Download RiskBench_Dataset here.

2. Download metadata.zip here.

3. Unzip model.zip, metadata.zip, RiskBench_Dataset

4. Install dependencies in your favorite environment.

   e.g. start by creating a new conda environment:

   conda create -n analysis_tool python=3.7
   conda activate analysis_tool
   cd ${TOOL_ROOT}
   
   conda  install pyqt
   pip3 install -r requirements.txt

Quantitative Results For Risk Object Identification (ROI)

1. Execute

   python ROI_tool.py --method ${MODEL} --data_type ${DATA_TYPE} --metadata_root ${METADATA_ROOT} --save_result --result_path ${ROI_PATH}

2. The results will be saved to ${ROI_PATH}/${MODEL}/${DATA_TYPE}.josn

Fine-grained Scenario-based Analysis

1. Execute

   python ROI_vis_tool.py --data_root ${DATASET_ROOT} --metadata_root ${METADATA_ROOT} --vis_result_path ${VIS_PATH}

2. Choose Model and Scenario Type

3. Check interest attributes

4. Click Filter Scenario

5. Choose one scenario

   • Click Generate Video to save the qualitative result in gif file
   • Click Generate JSON to save the quantitative result in JSON file
   • The results will be saved to ${VIS_PATH}/gif/${MODEL}/${DATA_TYPE} or {VIS_PATH}/json/${MODEL}/${DATA_TYPE}