Progressive Mirror Detection (PMD) Dataset

The PMD dataset is a benchmark for mirror surface detection introduced in the CVPR 2020 paper Progressive Mirror Detection.

Dataset Statistics

Split	Images	Mask	Edge
train	5,095	✓	✓
test	571	✓	—
total	5,666

Columns

Column	Type	Description
image_id	string	Original filename stem (e.g. 000000000164), unique within each split
image	Image (RGB JPEG)	Input scene photograph
mask	Image (grayscale PNG)	Binary mirror segmentation mask
edge	Image (grayscale PNG)	Mirror edge map (train only; None for test)

Usage

from datasets import load_dataset

ds = load_dataset("garrying/PMD")

sample = ds["train"][0]
sample["image_id"]  # original filename stem, e.g. "000000000164"
sample["image"]     # PIL Image (RGB)
sample["mask"]      # PIL Image (grayscale)
sample["edge"]      # PIL Image (grayscale, None for test)

Converting Back to Raw Files

A helper script parquet_to_raw.py is included to restore the original folder layout:

# download the script
huggingface-cli download garrying/PMD parquet_to_raw.py --repo-type dataset --local-dir .

# convert all splits
python parquet_to_raw.py --repo garrying/PMD --out PMD

Output layout:

PMD/
  train/  image/  mask/  edge/
  test/   image/  mask/

Pretrained Model

A pretrained PMDNet checkpoint is available at garrying/PMD-PMDNet.

License

This dataset is released under CC BY-NC 4.0. Please cite the paper below if you use this dataset in your work.

Citation

@INPROCEEDINGS{PMD:2020,
   Author    = {Jiaying Lin and Guodong Wang and Rynson W.H. Lau},
   Title     = {Progressive Mirror Detection},
   Booktitle = {Proc. CVPR},
   Year      = {2020}
}

Contact

csjylin@gmail.com