LIP Human Parsing
Pixel-level human body part segmentation (Gong et al., CVPR 2017). Parse human images into 19 semantic body-part and clothing labels including hair, face, arms, legs, and clothing items.
About this dataset
Human parsing is the task of segmenting a person in an image into semantic parts such as hair, face, arms, legs, and individual clothing items. The Look Into Person (LIP) benchmark was introduced by Ke Gong, Xiaodan Liang, Dongyu Zhang, Xiaohui Shen, and Liang Lin at CVPR 2017 to give the task a larger and more varied dataset than what existed before, covering a wider range of viewpoints, occlusions, and background complexity.
The images are real photographs of single people cropped from a broader collection, chosen to include difficult poses and partial occlusion. Each image is annotated at the pixel level, with every pixel assigned to one of the human part or clothing categories or to background.
LIP contains over 50,000 elaborately annotated images labeled with 19 semantic part categories (plus background). The authors report that this scale and diversity was a substantial step up from prior human-parsing datasets, and they use it to benchmark leading parsing methods and their own self-supervised, structure-sensitive approach.
The Potato config below reproduces this task with a multiselect over the 19 LIP body-part and clothing labels, plus radio questions for the primary pose and the degree of occlusion. It runs as an image-level part-presence and attribute pass, which is useful for triaging images or checking which parts are visible before a heavier pixel-level segmentation stage.
- Total images
- Over 50,000
- Semantic part labels
- 19
- Annotation
- Pixel-level part segmentation
- Task
- Single-person human parsing
- Authors
- Gong et al.
- Venue
- CVPR 2017
Configuration Fileconfig.yaml
This Potato config reproduces the annotation task. Save it as config.yaml and run potato start config.yaml to try it.
# LIP Human Parsing Configuration
# Based on Gong et al., CVPR 2017
annotation_task_name: "LIP Human Body Part Parsing"
task_dir: "."
data_files:
- "sample-data.json"
item_properties:
id_key: "id"
text_key: "image_url"
context_key: "context"
user_config:
allow_all_users: true
annotation_schemes:
- annotation_type: "multiselect"
name: "body_parts"
description: "Select all visible body parts"
labels:
- name: "hat"
tooltip: "Hat or head covering"
- name: "hair"
tooltip: "Hair"
- name: "face"
tooltip: "Face"
- name: "sunglasses"
tooltip: "Sunglasses or glasses"
- name: "upper_clothes"
tooltip: "Upper body clothing"
- name: "dress"
tooltip: "Dress"
- name: "coat"
tooltip: "Coat or jacket"
- name: "socks"
tooltip: "Socks"
- name: "pants"
tooltip: "Pants"
- name: "gloves"
tooltip: "Gloves"
- name: "scarf"
tooltip: "Scarf"
- name: "skirt"
tooltip: "Skirt"
- name: "jumpsuits"
tooltip: "Jumpsuits"
- name: "left_arm"
tooltip: "Left arm"
- name: "right_arm"
tooltip: "Right arm"
- name: "left_leg"
tooltip: "Left leg"
- name: "right_leg"
tooltip: "Right leg"
- name: "left_shoe"
tooltip: "Left shoe"
- name: "right_shoe"
tooltip: "Right shoe"
- annotation_type: "radio"
name: "pose_type"
description: "What is the primary pose?"
labels:
- name: "standing"
tooltip: "Person is standing"
- name: "sitting"
tooltip: "Person is sitting"
- name: "walking"
tooltip: "Person is walking"
- name: "other"
tooltip: "Other pose"
- annotation_type: "radio"
name: "occlusion"
description: "Are body parts occluded?"
labels:
- name: "no_occlusion"
tooltip: "All visible parts are clear"
- name: "partial_occlusion"
tooltip: "Some parts are partially hidden"
- name: "heavy_occlusion"
tooltip: "Significant occlusion"
interface_config:
item_display_format: "<img src='{{text}}' style='max-width:100%; max-height:500px;'/><br/><small>{{context}}</small>"
output_annotation_format: "json"
output_annotation_dir: "annotations"
Sample Datasample-data.json
[
{
"id": "lip_001",
"image_url": "https://upload.wikimedia.org/wikipedia/commons/thumb/c/c2/Aiga_toiletsq_men.svg/800px-Aiga_toiletsq_men.svg.png",
"context": "Parse this human image into semantic body parts. Label all visible body parts and clothing items."
},
{
"id": "lip_002",
"image_url": "https://upload.wikimedia.org/wikipedia/commons/thumb/e/e5/Red_coat.jpg/800px-Red_coat.jpg",
"context": "Segment body parts: face, hair, arms, legs, and clothing."
}
]
// ... and 1 more itemsGet This Design
Clone or download from the repository
Quick start:
git clone https://github.com/davidjurgens/potato-showcase.git cd potato-showcase/image/segmentation/lip-human-parsing potato start config.yaml
Dataset & paper
Gong, Liang, Zhang, Shen & Lin, CVPR 2017
Citation (BibTeX)
@inproceedings{gong2017look,
title = {Look into Person: Self-supervised Structure-sensitive Learning and a New Benchmark for Human Parsing},
author = {Gong, Ke and Liang, Xiaodan and Zhang, Dongyu and Shen, Xiaohui and Lin, Liang},
booktitle = {Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
pages = {932--940},
year = {2017}
}Details
Annotation Types
Domain
Use Cases
Tags
Found an issue or want to improve this design?
Open an IssueRelated Designs
ADE20K: Scene Parsing and Semantic Segmentation Dataset
ADE20K is a scene parsing dataset from MIT CSAIL (Zhou et al., CVPR 2017) with dense pixel labels for objects, stuff, and parts. This Potato config adapts it to an image-level category-tagging task over a curated subset of common ADE20K categories.
BDD100K: Diverse Driving Dataset for Autonomous Driving
BDD100K is a 100,000-video driving dataset from UC Berkeley (Yu et al., CVPR 2020) spanning detection, segmentation, lane marking, and tracking. This Potato config reproduces its driving-scene labeling tasks.
CUB-200-2011: Fine-Grained Bird Species Dataset
CUB-200-2011 is a fine-grained bird image dataset of 11,788 photos across 200 species, each labeled with parts, attributes, and a bounding box (Wah et al., 2011). This Potato config reproduces the labeling task.