#Dataset Assistance Tool

To facilitate your use, we also provide some dataset auxiliary tools. The following will introduce them to you based on the usage scenarios.

#Extract Camera Parameters

#Usage

Run tools/gen_camera_param_nusc.py to retrieve the intrinsic and extrinsic parameters of cameras from the nuScenes dataset.

#Parameters Introduction

#Usgae Example

Save the intrinsic and extrinsic parameter matrices of NuScenes cameras by city.

python3 tools/gen_camera_param_nusc.py --data-path /path/to/nuscenes --save-path ./camera_params_city --save-by-city --version v1.0-trainval

#Extract Model Input Reference Points

#Usage

Run tools/gen_reference_points_nusc.py to obtain the model input reference points from the nuScenes dataset.

#Parameters Introduction

#Usgae Example

python3 tools/gen_reference_points_nusc.py \
        --data-path /path/to/nuscenes \
        --save-path ./output/ref_points_city \
        --save-by-city True \
        --version v1.0-trainval \
        -c configs/bev/bevformer_tiny_resnet50_detection_nuscenes.py

#Compute Transformation Matrix

#Usage

Run tools/homography_generator.py to calculate the transformation matrix from ego to image.

#Parameters Introduction

#Usgae Example

python3 tools/homography_generator.py \
    --sensor2ego-translation ./data/s2e_translation.npy \
    --sensor2ego-rotation ./data/s2e_rotation.npy \
    --camera-intrinsic ./data/camera_intrinsic.npy \
    --save-path ./output/homography

#Compute Model Input Reference Points

#Usage

Run tools/reference_points_generator.py to calculate the model input reference points from the homography matrix.

#Parameters Introduction

#Usgae Example

python3 tools/reference_points_generator.py \
        --config ./configs/bev/bevformer_tiny_resnet50_detection_nuscenes.py \
        --homography ./output/homography.npy \
        --save-path ./output/reference_points

For a detailed introduction to the configuration file, please refer to the [Configuration] section [config configuration] (../config/keywords).