CoolFace
Datasetpublic

nobg/P3M-10K

P3M-10K P3M-10K (Privacy-Preserving Portrait Matting) is a large-scale portrait matting benchmark. It is redistributed here from the original release by JizhiziLi/P3M. If you use this dataset, please cite the original paper: Jizhizi Li, Sihan Ma, Xin Zhang, Dacheng Tao. "Privacy-Preserving Portrait Matting." ACM International Conference on Multimedia (ACM MM), 2021. Contents Each example is a portrait RGB image and its corresponding alpha matte: Column… See the full description on the dataset page: https://huggingface.co/datasets/nobg/P3M-10K.

sourceHugging Facemitupdated 2mo agoView on Hugging Face
0likes455downloads
Dataset Card

P3M-10K

P3M-10K (Privacy-Preserving Portrait Matting) is a large-scale portrait matting benchmark. It is redistributed here from the original release by JizhiziLi/P3M.

If you use this dataset, please cite the original paper:

Jizhizi Li, Sihan Ma, Xin Zhang, Dacheng Tao. "Privacy-Preserving Portrait Matting." ACM International Conference on Multimedia (ACM MM), 2021.

Contents

Each example is a portrait RGB image and its corresponding alpha matte:

ColumnTypeDescription
imageImageRGB portrait image
maskImageAlpha matte (soft foreground opacity)

Note: mask is the alpha matte, not a binary segmentation map.

Splits

SplitRowsImage sourceDescription
train9421blurred_image (privacy)Training set with face-blurred portraits
P3M_500_P500blurred_image (privacy)Validation set, Privacy (face-blurred) portraits
P3M_500_NP500original_imageValidation set, Non-Privacy (unblurred) portraits

The schema is identical across all three splits (image + mask).

Dropped columns

The original release also ships fg/ (foreground) and bg/ (background) for the training set, and trimap/ for the validation sets. These were dropped to keep a clean, consistent two-column schema (image, mask) across all splits.

License

Released under the MIT License, matching the original P3M-10K Dataset Release Agreement (MIT License). Please review and abide by the original agreement.