README.md

Our Dataset

Annotation format

We save the annotation in a JSON format, including 2D edge points and face loops by edge indices.

{
    'edges': [
        [...],  # edge 1
        [...],  # edge 2
        ...
    ],
    'faces_indices': [
        [...],  # face 1
        [...],  # face 2
        ...
    ],
}

Prepare Dataset

Here, we provide tools to filter and parse data in ABC dataset. Please download step, stat, obj, and feat.

Reorganize ABC dataset directory

Remove the middle level folder after unzipping ABC dataset.

python reorganize_dataset_dirs.py --root $ABC_ROOT_DIR

# Original ABC Dataset Structure
root
└── step
    └──00000050
       └── 00000050.step
# Reorganized ABC Dataset structure
root
└── step
    └── 00000050.step

Dataset Directory Structure

root
├── step
│   └── 00000050.step
├── json
│   └── 00000050.json
├── face_png
│   └── 00000050_{face_index}.png
├── face_svg
│   └── 00000050_{face_index}.svg
├── png
│   └── 00000050.png
└── svg
    └── 00000050.svg

Command Lines

Data Generation

In each model's config, we detail the specific options needed to generate dataset of the correct format.

# parse the entire ABC dataset
python dataset/prepare_data.py --root $ABC_ROOT_DIR --id_list dataset/dataset_gen_logs/filtered_id_list.json > dataset/dataset_gen_logs/error.txt
# parse a specific object (for debugging a single data)
python dataset/prepare_data.py --root $ABC_ROOT_DIR --name $8_DIGIT_ID

Dataset Filtering

Filter ABC objects by similarity

# 1. filter by topology similarity
python dataset/filters/filter_topology.py --root $ABC_ROOT_DIR
# 2. render the three views of the entire ABC dataset
python dataset/filters/3view_render.py --root $ABC_ROOT_DIR --id_list dataset/dataset_gen_logs/filtered_id_list.json > dataset/dataset_gen_logs/3view_error.txt
# 3. filter by three-view similarity
python dataset/filters/filter_3view.py --root $ABC_ROOT_DIR

Filter ABC objects by thickness

python dataset/filters/filter_thickness.py --root $ABC_ROOT_DIR --save_root $DIR_FOR_TEMP_DATA

Filter ABC objects by complexity

# By default, $MAX_FACE_SEQ = 128, $MAX_NUM_EDGE = 64
python dataset/filters/filter_length.py --root $ABC_ROOT_DIR --face_seq_max $MAX_FACE_SEQ --num_edge_max $MAX_NUM_EDGE

Filter Generated Co-edge Data by Face Encloseness

# Assume prepare_data.py has finished and all json files have generated
python dataset/tests/check_faces_enclosed.py --root $ABC_ROOT_DIR --tol 1e-4 --remove
# Regenerate train/valid/test splits
python dataset/prepare_data.py --root $ABC_ROOT_DIR --only_split