DeepFashion: Large-Scale Clothing Attributes Dataset
DeepFashion is a clothes recognition and retrieval dataset of 800,000+ images labeled with 50 categories, 1,000 attributes, and landmarks (Liu et al., CVPR 2016). This Potato config reproduces its category and attribute task.
About this dataset
DeepFashion is a large-scale clothes dataset built by Ziwei Liu, Ping Luo, Shi Qiu, Xiaogang Wang, and Xiaoou Tang at the Multimedia Laboratory, Chinese University of Hong Kong. It was published at CVPR 2016 to support robust clothes recognition and retrieval.
The dataset holds over 800,000 fashion images taken across store, street snapshot, and consumer settings, including more than 300,000 cross-pose and cross-domain image pairs. Each image is labeled with 50 clothing categories, 1,000 descriptive attributes, bounding boxes, and clothing landmarks.
The 1,000 attributes split into five groups: texture, fabric, part, shape, and style. Four benchmarks share the data: Category and Attribute Prediction, In-Shop Clothes Retrieval, Consumer-to-Shop Clothes Retrieval, and Fashion Landmark Detection. The Category and Attribute Prediction benchmark alone covers 289,222 images with 46 categories, 1,000 attributes, and 8 landmarks per image.
The Potato config below reproduces the category and attribute annotation task: an annotator picks one clothing category with a radio scheme and tags applicable attributes with a multiselect scheme.
- Total images
- 800,000+
- Clothing categories
- 50
- Descriptive attributes
- 1,000
- Cross-pose/cross-domain pairs
- 300,000+
- Benchmarks
- 4
- Category & Attribute benchmark
- 289,222 images, 8 landmarks
Configuration Fileconfig.yaml
This Potato config reproduces the annotation task. Save it as config.yaml and run potato start config.yaml to try it.
# DeepFashion Fine-Grained Fashion Classification Configuration
# Based on Liu et al., CVPR 2016
annotation_task_name: "DeepFashion Clothing Classification"
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: "radio"
name: "clothing_category"
description: "Select the clothing category"
labels:
- name: "upper_body"
tooltip: "Tops, shirts, jackets"
- name: "lower_body"
tooltip: "Pants, skirts, shorts"
- name: "full_body"
tooltip: "Dresses, jumpsuits"
- name: "outerwear"
tooltip: "Coats, jackets, blazers"
- annotation_type: "multiselect"
name: "texture_attributes"
description: "Select texture attributes"
labels:
- name: "solid"
tooltip: "Solid color"
- name: "striped"
tooltip: "Striped pattern"
- name: "floral"
tooltip: "Floral pattern"
- name: "plaid"
tooltip: "Plaid/checkered"
- name: "spotted"
tooltip: "Polka dots or spots"
- name: "graphic"
tooltip: "Graphics or prints"
- annotation_type: "multiselect"
name: "style_attributes"
description: "Select style attributes"
labels:
- name: "casual"
tooltip: "Casual style"
- name: "formal"
tooltip: "Formal/business"
- name: "sporty"
tooltip: "Athletic/sporty"
- name: "vintage"
tooltip: "Vintage/retro"
- name: "bohemian"
tooltip: "Bohemian style"
- name: "minimalist"
tooltip: "Minimalist design"
- annotation_type: "multiselect"
name: "fabric_attributes"
description: "Select fabric attributes"
labels:
- name: "cotton"
tooltip: "Cotton fabric"
- name: "denim"
tooltip: "Denim/jeans"
- name: "leather"
tooltip: "Leather"
- name: "silk"
tooltip: "Silk or satin"
- name: "wool"
tooltip: "Wool or knit"
- name: "lace"
tooltip: "Lace"
- annotation_type: "radio"
name: "sleeve_length"
description: "Sleeve length (if applicable)"
labels:
- name: "sleeveless"
tooltip: "No sleeves"
- name: "short_sleeve"
tooltip: "Short sleeves"
- name: "long_sleeve"
tooltip: "Long sleeves"
- name: "not_applicable"
tooltip: "Not applicable"
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": "fashion_001",
"image_url": "https://upload.wikimedia.org/wikipedia/commons/thumb/e/e5/Red_coat.jpg/800px-Red_coat.jpg",
"context": "Annotate this clothing item with category, texture, style, and fabric attributes."
},
{
"id": "fashion_002",
"image_url": "https://upload.wikimedia.org/wikipedia/commons/thumb/c/c2/Aiga_toiletsq_men.svg/800px-Aiga_toiletsq_men.svg.png",
"context": "Classify the clothing and select all applicable attributes."
}
]
// ... 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/specialized/deepfashion potato start config.yaml
Dataset & paper
Liu et al., CVPR 2016
Citation (BibTeX)
@inproceedings{liu2016deepfashion,
title = "{DeepFashion}: Powering Robust Clothes Recognition and Retrieval with Rich Annotations",
author = "Liu, Ziwei and Luo, Ping and Qiu, Shi and Wang, Xiaogang and Tang, Xiaoou",
booktitle = "Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition",
pages = "1096--1104",
year = "2016",
url = "https://openaccess.thecvf.com/content_cvpr_2016/html/Liu_DeepFashion_Powering_Robust_CVPR_2016_paper.html"
}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.