EasyPortrait - Face Parsing and Portrait Segmentation Dataset

We introduce a large-scale image dataset EasyPortrait for portrait segmentation and face parsing. Proposed dataset can be used in several tasks, such as background removal in conference applications, teeth whitening, face skin enhancement, red eye removal or eye colorization, and so on.

EasyPortrait dataset size is about 26GB, and it contains 20 000 RGB images (~17.5K FullHD images) with high quality annotated masks. This dataset is divided into training set, validation set and test set by subject user_id. The training set includes 14000 images, the validation set includes 2000 images, and the test set includes 4000 images.

https://github.com/hukenovs/easyportrait/blob/main/images/data.jpg?raw=true" alt="">

🔥 Changelog

• 2023/11/13: We release EasyPortrait 2.0. ✌️
  • 40,000 RGB images (~38.3K FullHD images)
  • Added diversity by region, race, human emotions and lighting conditions
  • The data was further cleared and new ones were added
  • Train/val/test split: (30,000) 75% / (4,000) 10% / (6,000) 15% by subject user_id
  • Multi-gpu training and testing
  • Added new models for face parsing and portrait segmentation
  • Dataset size is 91.78GB
  • 13,705 unique persons
• 2023/02/23: EasyPortrait (Initial Dataset) 💪
  • Dataset size is 26GB
  • 20,000 RGB images (~17.5K FullHD images) with 9 classes annotated
  • Train/val/test split: (14,000) 70% / (2,000) 10% / (4,000) 20% by subject user_id
  • 8,377 unique persons

Downloads

Link	Size
images	91.8 GB
annotations	657.1 MB
meta	1.9 MB
train set	68.3 GB
validation set	10.7 GB
test set	12.8 GB

Structure

.
├── images.zip
│  ├── train/     # Train set: 30k
│  ├── val/      # Validation set: 4k
│  ├── test/     # Test set: 6k
├── annotations.zip
│  ├── train/
│  ├── val/
│  ├── test/
├── meta.zip    # Meta-information (width, height, brightness, imhash, user_id)
...

Annotations

Annotations are presented as 2D-arrays, images in *.png format with several classes:

Index	Class
0	BACKGROUND
1	PERSON
2	SKIN
3	LEFT BROW
4	RIGHT_BROW
5	LEFT_EYE
6	RIGHT_EYE
7	LIPS
8	TEETH

Also, we provide some additional meta-information for dataset in annotations/meta.zip file:

	attachment_id	user_id	data_hash	width	height	brightness	train	test	valid
0	de81cc1c-...	1b...	e8f...	1440	1920	136	True	False	False
1	3c0cec5a-...	64...	df5...	1440	1920	148	False	False	True
2	d17ca986-...	cf...	a69...	1920	1080	140	False	True	False

where: - attachment_id - image file name without extension - user_id - unique anonymized user ID - data_hash - image hash by using Perceptual hashing - width - image width - height - image height - brightness - image brightness - train, test, valid are the binary columns for train / test / val subsets respectively

Models

We provide some pre-trained models as the baseline for portrait segmentation and face parsing. We use mean Intersection over Union (mIoU) as the main metric.

Pretrained models are available in github repo!

Links

• Github
• arXiv
• Paperswithcode
• Habr

Authors and Credits

• Alexander Kapitanov
• Karina Kvanchiani
• Elizaveta Petrova
• [K...