auto_labeling_2d

auto_label_2d

1. Setup environment

• Support priority: Tier B
• See setting environemnt

2. command

(TBD) 2.1 T4dataset to info file

This tool can be used for non-annotated T4dataset and annotated T4dataset.

• Set T4dataset

- data/t4dataset/
  - pseudo_xx1_2d/
    - 0/
    - 1/

• Make the info file from T4dataset with offline model.
  • The info file contain the confidence information.

tools/t4dataset_pseudo_label_2d/t4dataset_pseudo_label.py --input {path to non-annotated T4dataset} --config {config_file} --ckpt {ckpt_file}

• As a result, the data is as below.

- data/t4dataset/
  - pseudo_xx1_2d/
    - 0/
    - 1/
    - ..
  - info/
    - pseudo_label_infos_2d.pkl

• The information of pseudo_label_infos_2d.pkl

- info\['cam_instances'\]\['CAM_FRONT'\]\[i\]\['bbox_label'\]: Label of instance.
- info\['cam_instances'\]\['CAM_FRONT'\]\[i\]\['bbox_label_3d'\]: Label of instance.
- info\['cam_instances'\]\['CAM_FRONT'\]\[i\]\['bbox'\]: 2D bounding box annotation (exterior rectangle of the projected 3D box), a list arrange as \[x1, y1, x2, y2\].
- info\['cam_instances'\]\['CAM_FRONT'\]\[i\]\['center_2d'\]: Projected center location on the image, a list has shape (2,), .

(TBD) 2.2 info file to pseudo-label T4dataset

• Set non-annotated T4dataset and info file

- data/t4dataset/
  - pseudo_xx1_2d/
    - 0/
      - annotation/
        - ..
    - 1/
    - ..
  - info/
    - pseudo_label_infos_2d.pkl

• Run script

tools/t4dataset_pseudo_label_2d/create_pseudo_t4dataset.py --input {path to non-annotated T4dataset} --threshold 0.3

• As a result, pseudo-label T4dataset is made as below.

- data/t4dataset/
  - pseudo_xx1_2d/
    - 0/
      - annotation/
        - sample.json
        - ..
    - 1/
    - ..