PPM-100 Dataset

Estimated reading: 2 minutes

Visualization of the PPM-100 dataset in the Deep Lake UI

PPM-100 dataset

What is PPM-100 Dataset?

The PPM-100 (Photography Portrait Matting 100) Dataset is a portrait matting benchmark dataset that contains carefully collected images with respective masks. All the images are diverse and they contain full/half-body poses. These images use their original background, and the image resolution is between 1080p and 4k.

Download PPM-100 Dataset in Python

Instead of downloading the PPM-100 dataset in Python, you can effortlessly load it in Python via our Deep Lake open-source with just one line of code.

Load PPM-100 Dataset Subset in Python

				
					import deeplake
ds = deeplake.load('hub://activeloop/ppm100')

PPM-100 Dataset Structure

Data Fields

image: tensor containing the image.
masks: tensor containing the image

How to use PPM-100 Dataset with PyTorch and TensorFlow in Python

Train a model on PPM-100 dataset with PyTorch in Python

Let’s use Deep Lake built-in PyTorch one-line dataloader to connect the data to the compute:

				
					dataloader = ds.pytorch(num_workers=0, batch_size=4, shuffle=False)

Train a model on PPM-100 dataset with TensorFlow in Python

				
					dataloader = ds.tensorflow()

Additional Information about PPM-100 Dataset

PPM-100 Dataset Description

Repository: https://github.com/ZHKKKe/PPM
Paper: https://arxiv.org/pdf/2011.11961.pdf
Point of Contact: Zhanghan Ke – [email protected]

PPM-100 Dataset Licensing Information

All original portrait images in PPM are from Flickr and constrained by Flickr Creative Commons License (Commercial use & mods allowed).

All annotated alpha mattes in PPM are released under the Creative Commons Attribution NonCommercial ShareAlike 4.0 license.

PPM-100 Dataset Citation Information

				
					@article{ke2020green,
  title={Is a Green Screen Really Necessary for Real-Time Portrait Matting?},
  author={Ke, Zhanghan and Li, Kaican and Zhou, Yurou and Wu, Qiuhua and Mao, Xiangyu and Yan, Qiong and Lau, Rynson WH},
  journal={arXiv preprint arXiv:2011.11961},
  year={2020}
}