This is a summary. We reviewed all (to our best knowledge) fashion-related papers in the past decade and recorded the datasets had been used. The numbers to describe the dataset is faithfully followed its original paper. The webpage is organized as:
The sections are defined according to the types of data, e.g. if you want clothing segmentation information, you can see Section 0. parsing to find annotated data.
If you want to obtain some attributes, you can see Section 1. attribute. We present the type of attributes (e.g. brand, review, style, comment, neckline, color etc), the number of images, potential tasks, type of images (e.g. product image, model image, which view, etc) for a quick check.
๐ท means the dataset can be found. If you find the dataset helpful, please kindly cite it in your paper ("bibtex" is offered for your convenience)~
Meanwhile, for consistency, we uniformed the words decribe the fashion concept.
- ย Silhouette (shape, cut, fit): the shape of a garment, e.g. H line, A line etc;
- ย Material (fabric): the material made a garment, e.g. chiffon, lace etc;
- ย Print (pattern, texture): the surface design of a garment, e.g. checks, dotted etc;
- ย Neckline (collar shape, collar): the design in the neck region of a garment, e.g. V-neck, lapel etc;
- ย Design details (structures, style): designs which can be used in anywhere of a garment, e.g. frilly, ruffled etc;
- ย Opening (cloth button, fastening): the way designed in the opening of a garment e.g. button, zipped etc;
- ย Category (type): type of a garment, e.g. dress, top etc;
- ย Sub-category: fine-grained type of a garment, e.g. wedding dress, T-shire etc;
- ย Styles (looks): the expressed feeling of a garment of an outfit, e.g. lovely, casual etc;
- ย Gender (Persons): Men's wear, women's wear (child, boy, female) etc;
- ย Design Attributes: the attributes used in the process of garment design, e.g. shirt cuff, shirt collar etc;
- ย Retail Attributes: the attributes used in the process of retail, e.g. parka, windbreaker etc.
If you have problems with a specific dataset shows below, please kindly contact its authors. For a quick check, you can also see my own memo version ยฉ
- 0. ย Parsing
- 1. ย Keypoints
- 2. ย Attribute
- 3. ย Outfit
- 4. ย Generation
- 5. ย Others
- Other Sources
- Acknowledge
For semantic segmentation, object detection, instance segmentation, polygon detection, and etc.
๐ Fashionista 2012
(1) 158,235 fashion photos with associated text annotations (tags, comments, and links).
(2) The tags are noisy or incomplete.
(3) 685 photos with good visibility of the full-body with pose annotations for the usual 14 body parts.
(4) There are totally 56 labels (53 category or sub-category labels, and additional labels for hair, skin, and null (background).
[homepage] [pdf] ๐ท
@inproceedings{yamaguchi2012parsing,
title={Parsing clothing in fashion photographs},
author={Yamaguchi, Kota and Kiapour, M Hadi and Ortiz, Luis E and Berg, Tamara L},
booktitle={2012 IEEE Conference on Computer Vision and Pattern Recognition},
pages={3570--3577},
year={2012},
organization={IEEE}
}๐ Paperdoll 2013
(1) Over 1 million pictures from chictopia.com with associated metadata tags i.e. color, clothing item, or occasion.
(2) 339,797 pictures weakly annotated with clothing items and estimated pose.
(3) 685 fully parsed images .
[homepage] [pdf] ๐ท
@inproceedings{yamaguchi2013paper,
title={Paper doll parsing: Retrieving similar styles to parse clothing items},
author={Yamaguchi, Kota and Hadi Kiapour, M and Berg, Tamara L},
booktitle={Proceedings of the IEEE international conference on computer vision},
pages={3519--3526},
year={2013}
}๐ CFPD 2013
(1) 97,490 images with 292,541 tags from Chictopia.com.
(2) 2,682 images in total, and all the pixels in the images are annotated with both color labels (13) and category labels (23).
(3) Weakly supervised setting, where only image-level tags are available in the training phase.
[homepage] [pdf] [github1] [github2] ๐ท
@article{liu2013fashion,
title={Fashion parsing with weak color-category labels},
author={Liu, Si and Feng, Jiashi and Domokos, Csaba and Xu, Hui and Huang, Junshi and Hu, Zhenzhen and Yan, Shuicheng},
journal={IEEE Transactions on Multimedia},
volume={16},
number={1},
pages={253--265},
year={2013},
publisher={IEEE}
}๐ CCP 2013
(1) It consisting of 2098 high-resolution street fashion photos.
(2) More than 1,000 images are annotated with superpixel-level labeling with a total of 57 tags.
(3) Cross-scenario image pairs, which include about 10,000 product photos and user's photos image pairs.
(4) Each image has 124 fine-grained semantic attributes.
(5) 20 categories, 56 colors, 6 clothing length, 10 silhouette, 25 necklines, and 7 sleeve length.
[homepage] [pdf] [github] ๐ท
@inproceedings{yang2014clothing,
title={Clothing Co-Parsing by Joint Image Segmentation and Labeling},
author={Yang, Wei and Luo, Ping and Lin, Liang}
booktitle={Computer Vision and Pattern Recognition (CVPR), 2014 IEEE Conference on},
year={2013},
organization={IEEE}
}๐ HCP 2015
(1) 7,700 images in total.
(2) Combined Fashionista (685), CFPD (2,682), Daily Photos dataset (2,500).
(3) Crawl another 1,833 challenging images ๏ผe.g. sitting or occlusion) annotate pixel-level labels.
(4) 18 categories of labels, e.g. face, sunglass, hat, scarf etc.
[homepage] [pdf] ๐ท
@article{liang2015deep,
title={Deep human parsing with active template regression},
author={Liang, Xiaodan and Liu, Si and Shen, Xiaohui and Yang, Jianchao and Liu, Luoqi and Dong, Jian and Lin, Liang and Yan, Shuicheng},
journal={IEEE transactions on pattern analysis and machine intelligence},
volume={37},
number={12},
pages={2402--2414},
year={2015},
publisher={IEEE}
}๐ Fashion Icon 2015
(1) Video dataset and Fashion Icon (FI) image dataset.
(2) Video dataset contains 1, 500 videos.
(3) FI image dataset contains 1, 082 images, 18 categories.
[homepage] [pdf]
@article{liu2015fashion,
title={Fashion parsing with video context},
author={Liu, Si and Liang, Xiaodan and Liu, Luoqi and Lu, Ke and Lin, Liang and Cao, Xiaochun and Yan, Shuicheng},
journal={IEEE Transactions on Multimedia},
volume={17},
number={8},
pages={1347--1358},
year={2015},
publisher={IEEE}
}๐ Refined Fashionista 2017
(1) Reduces the number of clothing categories from 56 to 25 essential labels.
(2) Manually annotated all the 685 images in the Fashionista dataset.
[homepage] [pdf] [github] ๐ท
@article{tangseng2017looking,
title={Looking at outfit to parse clothing},
author={Tangseng, Pongsate and Wu, Zhipeng and Yamaguchi, Kota},
journal={arXiv preprint arXiv:1703.01386},
year={2017}
}๐ FASHION8 2018
(1) 9,339 fashion images from 8 continuous years are collected.
(2) With human-annotated foreground masks.
[homepage] [pdf]
@article{zhang2018fusing,
title={Fusing Hierarchical Convolutional Features for Human Body Segmentation and Clothing Fashion Classification},
author={Zhang, Zheng and Song, Chengfang and Zou, Qin},
journal={arXiv preprint arXiv:1803.03415},
year={2018}
}๐ ModaNet 2018
(1) 55, 176 street images, fully annotated with polygons (bounding box, segmentation mask).
(2) Based on 1 million weakly annotated street images in Paperdoll.
(3) 13 categories annotated (e.g. bag, belt, boots).
[homepage] [pdf] [github] ๐ท
@inproceedings{zheng2018modanet,
title={Modanet: A large-scale street fashion dataset with polygon annotations},
author={Zheng, Shuai and Yang, Fan and Kiapour, M Hadi and Piramuthu, Robinson},
booktitle={Proceedings of the 26th ACM international conference on Multimedia},
pages={1670--1678},
year={2018}
}For keypoint detection, landmark detection, pose estimation and etc.
๐ FLD 2016
(1) Over 120K images.
(2) Each image is correctly labeled with 8 fashion landmarks along with their visibility.
(3) Different types of clothing items, including upper/lower/full-body clothes.
(4) Different subsets, including normal/medium/large poses and medium/large scales.
[homepage] [pdf] [github] ๐ท
@inproceedings{liu2016fashionlandmark,
author = {Ziwei Liu, Sijie Yan, Ping Luo, Xiaogang Wang, and Xiaoou Tang},
title = {Fashion Landmark Detection in the Wild},
booktitle = {European Conference on Computer Vision (ECCV)},
month = {October},
year = {2016}
}๐ FASHIONAI Keypoint 2018
(1) 24 key points in 324k image (including armpit, crotch keypoints).
[homepage] [pdf] [TIANCHI] ๐ท
๐ You may kindly obtain the data by logging in to the TIANCHI platform (Choosing the international version, then register a personal account and sign the agreement. You will be supposed to access the data successfully).
@inproceedings{zou2019fashionai,
title={FashionAI: A Hierarchical Dataset for Fashion Understanding},
author={Zou, Xingxing and Kong, Xiangheng and Wong, Waikeung and Wang, Congde and Liu, Yuguang and Cao, Yang},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops},
pages={0--0},
year={2019}
}๐ DeepFashion2 2019
(1) 801K clothing items annotated with style, scale, viewpoint, occlusion, bounding box, dense landmarks, masks.
(2) 873K Commercial-Consumer clothes pairs.
(3) 13 different definitions of landmarks and poses for 13 different categories.
[homepage] [pdf] [github] ๐ท
@article{DeepFashion2,
author = {Yuying Ge and Ruimao Zhang and Lingyun Wu and Xiaogang Wang and Xiaoou Tang and Ping Luo},
title={A Versatile Benchmark for Detection, Pose Estimation, Segmentation and Re-Identification of Clothing Images},
journal={CVPR},
year={2019}
}For style analysis, attribute recognition, trend anaylsis, style anaylsis, multi-task learning, consumer-to-shop clothes retrieval, in-shop clothes retrieval and etc.
๐ Apparel Style 2012
(1) 15 categories (e.g. Long dress, Coat, Jacket etc).
(2) Over 80, 000 images.
(3) Attributes including colors(13), print(15), material(8), design details(4), styles(4+21), gender(5), sleeve length(3).
[homepage] [pdf] ๐ท
@inproceedings{bossard2012apparel,
title={Apparel classification with style},
author={Bossard, Lukas and Dantone, Matthias and Leistner, Christian and Wengert, Christian and Quack, Till and Van Gool, Luc},
booktitle={Asian conference on computer vision},
pages={321--335},
year={2012},
organization={Springer}
}๐ WFC 2013
(1) Womenโs Coat Dataset contains 2,092 images total with manuly annotated labels.
(2) 12 coat/jacket categories: cape, military, motorcycle, peacoat, poncho, puffer, trench, etc.
(3) 27 (Material(5), Fastener(4), Fastener style(3), clothing length(3), silhouette(2), pocket(2), neckline(8)) binary attributes.
[homepage] [pdf]
@inproceedings{di2013style,
title={Style finder: Fine-grained clothing style detection and retrieval},
author={Di, Wei and Wah, Catherine and Bhardwaj, Anurag and Piramuthu, Robinson and Sundaresan, Neel},
booktitle={Proceedings of the IEEE Conference on computer vision and pattern recognition workshops},
pages={8--13},
year={2013}
}๐ ZOZOTOWN 2013
(1) 12,719 photos on ZOZOTOWN.
(2) Have several meta-data, gender information, wearing items information, and photo date.
[homepage] [pdf]
@inproceedings{miura2013snapper,
title={SNAPPER: fashion coordinate image retrieval system},
author={Miura, Shinya and Yamasaki, Toshihiko and Aizawa, Kiyoharu},
booktitle={2013 International Conference on Signal-Image Technology \& Internet-Based Systems},
pages={784--789},
year={2013},
organization={IEEE}
}๐ Fashion136K 2013
(1) 135,893 street fashion images with annotations by fashionistas, brand, demographics.
(2) 3-4 annotations per image.
[homepage] [pdf]
@inproceedings{jagadeesh2014large,
title={Large scale visual recommendations from street fashion images},
author={Jagadeesh, Vignesh and Piramuthu, Robinson and Bhardwaj, Anurag and Di, Wei and Sundaresan, Neel},
booktitle={Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining},
pages={1925--1934},
year={2014}
}๐ UT-Zap50K 2014
(1) 50,000 shoe images with fine-grained attributes.
(2) 4 relative attributes: โopenโ, โpointy at the toeโ, โsportyโ, and โcomfortable".
(3) 12,000 total pairs, 3,000 per attribute.
[homepage] [pdf] ๐ท
@InProceedings{finegrained,
author = {A. Yu and K. Grauman},
title = {Fine-Grained Visual Comparisons with Local Learning},
booktitle = {Computer Vision and Pattern Recognition (CVPR)},
month = {Jun},
year = {2014}
}๐ Fashion 10000 2014
(1) 32,398 photos, with their associated metadata, distributed in 262 di๏ฌerent fashion categories.
[homepage] [pdf] ๐ท
@inproceedings{loni2014fashion,
title={Fashion 10000: an enriched social image dataset for fashion and clothing},
author={Loni, Babak and Cheung, Lei Yen and Riegler, Michael and Bozzon, Alessandro and Gottlieb, Luke and Larson, Martha},
booktitle={Proceedings of the 5th ACM Multimedia Systems Conference},
pages={41--46},
year={2014}
}๐ Hipster Wars 2014
(1) 1,893 images labeled with 5 style categories: hipster, bohemian, pinup, preppy, and goth.
[homepage] [pdf]
@inproceedings{kiapour2014hipster,
title={Hipster wars: Discovering elements of fashion styles},
author={Kiapour, M Hadi and Yamaguchi, Kota and Berg, Alexander C and Berg, Tamara L},
booktitle={European conference on computer vision},
pages={472--488},
year={2014},
organization={Springer}
}๐ Fashion144K 2015
(1) 144,169 user posts containing diverse images, textual, and meta information.
(2) Labels like location, comments, votes etc.
[homepage] [pdf] ๐ท
@InProceedings{SimoSerraCVPR2015,
author = {Edgar Simo-Serra and Sanja Fidler and Francesc Moreno-Noguer and Raquel Urtasun},
title = {{Neuroaesthetics in Fashion: Modeling the Perception of Fashionability}},
booktitle = "Proceedings of the Conference on Computer Vision and Pattern Recognition (CVPR)",
year = 2015,
}๐ WITB(Exact Street2Shop) 2015
(1) 404,683 shop photos and 20,357 street photos.
(2) Providing a total of 39,479 clothing item matches.
[homepage] [pdf] ๐ท
@inproceedings{hadi2015buy,
title={Where to buy it: Matching street clothing photos in online shops},
author={Hadi Kiapour, M and Han, Xufeng and Lazebnik, Svetlana and Berg, Alexander C and Berg, Tamara L},
booktitle={Proceedings of the IEEE international conference on computer vision},
pages={3343--3351},
year={2015}
}๐ DARN 2015
(1) 453,983 online upper-clothing images with 179 attributes in high-resolution.
(2) Each image contains a single frontal-view person.
(3) Opening(12), Category(20), Color(56), Cloth length(6), Print(27), Silhouette(10), Neckline(25), Sleeve length(7), Sleeve shape(16).
[homepage] [pdf]
@inproceedings{huang2015cross,
title={Cross-domain image retrieval with a dual attribute-aware ranking network},
author={Huang, Junshi and Feris, Rogerio S and Chen, Qiang and Yan, Shuicheng},
booktitle={Proceedings of the IEEE international conference on computer vision},
pages={1062--1070},
year={2015}
}๐ Clothing1M 2015
(1) 1, 000, 000 clothing images with 14 class labels: T-shirt, Shirt, Knitwear, Chiffon, Sweater, Hoodie etc.
(2) Each image is automatically assigned with a noisy label according to the keywords in its surrounding text.
(3) Manually refine 72, 409 image labels, which constitute a clean sub-dataset.
[homepage] [pdf] [github] ๐ท
@inproceedings{xiao2015learning,
title={Learning from Massive Noisy Labeled Data for Image Classification},
author={Xiao, Tong and Xia, Tian and Yang, Yi and Huang, Chang and Wang, Xiaogang},
booktitle={CVPR},
year={2015}
}๐ YahooClothing 2015
(1) 161,234 fashion images in the Yahoo shopping dataset.
(2) It labeled with category, gender, and sub-category(15) such as Top, Dress, Coat etc.
[homepage] [pdf]
@inproceedings{lin2015rapid,
title={Rapid clothing retrieval via deep learning of binary codes and hierarchical search},
author={Lin, Kevin and Yang, Huei-Fang and Liu, Kuan-Hsien and Hsiao, Jen-Hao and Chen, Chu-Song},
booktitle={Proceedings of the 5th ACM on International Conference on Multimedia Retrieval},
pages={499--502},
year={2015}
}๐ Chitopia 2015
(1) Chictopia dataset has 26,8124 usable images, each image has 2 clothing-keywords under 18 categories.
(2) Dress dataset consists of 712 images with total of 58 attributes.
[homepage] [pdf]
@inproceedings{yamaguchi2015mix,
title={Mix and Match: Joint Model for Clothing and Attribute Recognition.},
author={Yamaguchi, Kota and Okatani, Takayuki and Sudo, Kyoko and Murasaki, Kazuhiko and Taniguchi, Yukinobu},
booktitle={BMVC},
volume={1},
number={2},
pages={4},
year={2015}
}๐ Etsy | Wear 2016
(1) Etsy dataset has 173,175 clothing products.
(2) Wear dataset has 212,129 images associated shots from different views, list of items, blog text, tags, and other metadata.
[homepage] [pdf] ๐ท
@inproceedings{vittayakorn2016automatic,
title={Automatic attribute discovery with neural activations},
author={Vittayakorn, Sirion and Umeda, Takayuki and Murasaki, Kazuhiko and Sudo, Kyoko and Okatani, Takayuki and Yamaguchi, Kota},
booktitle={European Conference on Computer Vision},
pages={252--268},
year={2016},
organization={Springer}
}๐ DeepFashion 2016
(1) Over 800,000 images, which are richly annotated with massive attributes, clothing landmarks, and correspondence of images.
(2) 50 fine-grained categories and 1, 000 descriptive attributes.
(3) Attributes including Category, Print, Material, Silhouette, Part, Style.
[homepage] [pdf] ๐ท
@inproceedings{liuLQWTcvpr16DeepFashion,
author = {Liu, Ziwei and Luo, Ping and Qiu, Shi and Wang, Xiaogang and Tang, Xiaoou},
title = {DeepFashion: Powering Robust Clothes Recognition and Retrieval with Rich Annotations},
booktitle = {Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2016}
}๐ MVC 2016
(1) 37,499 items and 161,638 clothing images, 264 attributes(gender, category, sub-catgegory, design attributes).
(2) Multi-view(front, back, right, left) with high-resolution.
[homepage] [pdf] [github] ๐ท
@inproceedings{liu2016mvc,
title={Mvc: A dataset for view-invariant clothing retrieval and attribute prediction},
author={Liu, Kuan-Hsien and Chen, Ting-Yen and Chen, Chu-Song},
booktitle={Proceedings of the 2016 ACM on International Conference on Multimedia Retrieval},
pages={313--316},
year={2016}
}๐ StreetStyle27K 2017
(1) 27K images, each with 12 clothing attributes.
(2) A first-of-its-kind analysis of global and per-city fashion choices and trends.
(3) 7 binary attributes(Wearing Jacket, Wearing Scarf etc), 13 colors, 7 categories, 3 sleeve length, 3 neckline, 6 print.
[homepage] [pdf] ๐ท
@article{StreetStyle2017,
title={{StreetStyle}: {E}xploring world-wide clothing styles from millions of photos},
author={Kevin Matzen and Kavita Bala and Noah Snavely},
journal={arXiv preprint arXiv:1706.01869},
year={2017}
}๐ Fashionstyle14 2017
(1) 13,126 images labeled with 14 different fashion styles (conservative, dressy, ethnic, fairy, feminine etc).
[homepage] [pdf] ๐ท
@inproceedings{takagi2017makes,
title={What makes a style: Experimental analysis of fashion prediction},
author={Takagi, Moeko and Simo-Serra, Edgar and Iizuka, Satoshi and Ishikawa, Hiroshi},
booktitle={Proceedings of the IEEE International Conference on Computer Vision Workshops},
pages={2247--2253},
year={2017}
}๐ Fashion200K 2017
(1) Over 200,000 images of five categories (dress, top, pants, skirt, and jacket) and their product descriptions.
(2) After dealing with product description, 4,404 attributes(words) have remained.
[homepage] [pdf] [github] ๐ท
@inproceedings{han2017automatic,
title = {Automatic Spatially-aware Fashion Concept Discovery},
author = {Han, Xintong and Wu, Zuxuan and Huang, Phoenix X. and Zhang, Xiao and Zhu, Menglong and Li, Yuan and Zhao, Yang and Davis, Larry S.},
booktitle = {ICCV},
year = {2017},
}๐ Fashion550K 2017
(1) weakly-labeled image dataset consists of 550,661 images that includes 5,300 human-annotated images.
(2) 66 binary labels, 26 colors, 22 categories, 7 shoes, 11 accessories.
[homepage] [pdf] ๐ท
@InProceedings{InoueICCVW2017,
author = {Naoto Inoue and Edgar Simo-Serra and Toshihiko Yamasaki and Hiroshi Ishikawa},
title = {{Multi-Label Fashion Image Classification with Minimal Human Supervision}},
booktitle = "Proceedings of the International Conference on Computer Vision Workshops (ICCVW)",
year = 2017,
}๐ Amazon Dress
(1) 53 689 images of dresses and their product descriptions.
(2) Different categories, such as bridesmaid, casual, mother of the bride, night out and cocktail, and wedding.
(3) The product descriptions consist of the surrounding natural language text, like the title, features, and editorial content.
[homepage] [pdf]
@inproceedings{laenen2017cross,
title={Cross-modal search for fashion attributes},
author={Laenen, Katrien and Zoghbi, Susana and Moens, Marie-Francine},
booktitle={Proceedings of the KDD 2017 Workshop on Machine Learning Meets Fashion},
volume={2017},
pages={1--10},
year={2017},
organization={ACM}
}๐ SFS 2017
(1) 293,105 posts, each image consists of weakly-labeled multi-task ground-truth.
(2) Labels including season, occasion, fashion style, garment categories, geographical and year information.
[homepage] [pdf] ๐ท
@inproceedings{gu2017understanding,
title={Understanding fashion trends from street photos via neighbor-constrained embedding learning},
author={Gu, Xiaoling and Wong, Yongkang and Peng, Pai and Shou, Lidan and Chen, Gang and Kankanhalli, Mohan S},
booktitle={Proceedings of the 25th ACM international conference on Multimedia},
pages={190--198},
year={2017}
}๐ Video2Shop 2017
(1) 26,352 clothing trajectories extracted from 526 videos and 85,677 clothing shopping images.
(2) 14 categories of clothes are manually labeled.
[homepage] [pdf]
@inproceedings{cheng2017video2shop,
title={Video2shop: Exact matching clothes in videos to online shopping images},
author={Cheng, Zhi-Qi and Wu, Xiao and Liu, Yang and Hua, Xian-Sheng},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
pages={4048--4056},
year={2017}
}๐ RFS | PFS 2018
[homepage] [[pdf]]
@article{gu2018multi,
title={Multi-Modal and Multi-Domain Embedding Learning for Fashion Retrieval and Analysis},
author={Gu, Xiaoling and Wong, Yongkang and Shou, Lidan and Peng, Pai and Chen, Gang and Kankanhalli, Mohan S},
journal={IEEE Transactions on Multimedia},
volume={21},
number={6},
pages={1524--1537},
year={2018},
publisher={IEEE}
}๐ BrandFashion 2018
(1) 10K clothing images with distinctive logos from 15 brands.
(2) Label with 16 clothing categories and 32 semantic attributes.
[homepage] [pdf]
@inproceedings{manandhar2018tiered,
title={Tiered Deep Similarity Search for Fashion},
author={Manandhar, Dipu and Bastan, Muhammet and Yap, Kim-Hui},
booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},
pages={0--0},
year={2018}
}๐ FashionBrand 2018
(1) 3,828,735 clothing product images from 1,219 brands.
(2) Attributes including 5 categories, 13 sub-categories.
[homepage] [pdf]
@inproceedings{hadi2018brand,
title={Brand> Logo: Visual Analysis of Fashion Brands},
author={Hadi Kiapour, M and Piramuthu, Robinson},
booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},
pages={0--0},
year={2018}
}๐ Fashion60 2018
(1) 539,704 images for training and 30, 000 images for testing.
(2) 60 fine-grained fashion attributes categorized into 5 coarse-grained groups.
[homepage] [pdf]
@inproceedings{kuang2018ontology,
title={Ontology-driven hierarchical deep learning for fashion recognition},
author={Kuang, Zhenzhong and Yu, Jun and Yu, Zhou and Fan, Jianping},
booktitle={2018 IEEE Conference on Multimedia Information Processing and Retrieval (MIPR)},
pages={19--24},
year={2018},
organization={IEEE}
}๐ X-domain 2018
(1) 245,467 shop images. Each image is annotated by 9 attribute labels.
(2) Imbalanced with an imbalance-ratio of 1:4,162 (20 : 204,177).
(3) 178 distinctive attributes over the 9 labels, including 6 sleeve-length, 55 colors.
[homepage] [pdf]
@article{dong2018imbalanced,
title={Imbalanced deep learning by minority class incremental rectification},
author={Dong, Qi and Gong, Shaogang and Zhu, Xiatian},
journal={IEEE transactions on pattern analysis and machine intelligence},
volume={41},
number={6},
pages={1367--1381},
year={2018},
publisher={IEEE}
}๐ Women | Men Video 2018
(1) Video dataset for unsupervised learning.
(2) Two new clothing image datasets are annotated with 10 pre-defined clothing attributes.
(3) 18,737 woman clothing videos and 21,224 man clothing videos.
[homepage] [pdf]
@article{zhang2018watch,
title={Watch fashion shows to tell clothing attributes},
author={Zhang, Sanyi and Liu, Si and Cao, Xiaochun and Song, Zhanjie and Zhou, Jie},
journal={Neurocomputing},
volume={282},
pages={98--110},
year={2018},
publisher={Elsevier}
}๐ FASHIONAI Attributes 2018
(1) 324k images with 245 labels that cover 6 categories of womenโs clothing, and a total of 41 subcategories (single labeled).
[homepage] [pdf] [TIANCHI] ๐ท
๐ You may kindly obtain the data by logging in to the TIANCHI platform (Choosing the international version, then register a personal account and sign the agreement. You will be supposed to access the data successfully).
@inproceedings{zou2019fashionai,
title={FashionAI: A Hierarchical Dataset for Fashion Understanding},
author={Zou, Xingxing and Kong, Xiangheng and Wong, Waikeung and Wang, Congde and Liu, Yuguang and Cao, Yang},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops},
pages={0--0},
year={2019}
}๐ Feidegger 2018
(1) 8,700 fashion items, each with a high-resolution image and 5 independently collected textual descriptions. (2) Restrict the domain to images of dresses, and German-language visual descriptions.
@inproceedings{lefakis2018feidegger,
title={FEIDEGGER: A Multi-modal Corpus of Fashion Images and Descriptions in German},
author={Lefakis, Leonidas and Akbik, Alan and Vollgraf, Roland},
booktitle={Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)},
year={2018}
}๐ Studio2Shop 2018
(1) 1.15 million query images with neutral backgrounds. (2) Category(7), Color(82), Print(19), Cloth length (12), sleeve length (9), shirt collar(27), Neckline (12), Material (14), Trouser rise (3).
[homepage] [pdf]
@article{lasserre2018studio2shop,
title={Studio2shop: from studio photo shoots to fashion articles},
author={Lasserre, Julia and Rasch, Katharina and Vollgraf, Roland},
journal={arXiv preprint arXiv:1807.00556},
year={2018}
}๐ Shopping 100k 2018
(1) 101,021 images that consist of pure clothing items.
(2) Category(16), Neckline(17+11), Color(19), Material(14), Opening(9), Silhouette(15), Gender(2), Cloth length(7), Print(15), Pocket(7), Sleeve length(9), Sport(15).
[homepage] [pdf]
@inproceedings{ak2018efficient,
title={Efficient multi-attribute similarity learning towards attribute-based fashion search},
author={Ak, Kenan E and Lim, Joo Hwee and Tham, Jo Yew and Kassim, Ashraf A},
booktitle={2018 IEEE Winter Conference on Applications of Computer Vision (WACV)},
pages={1671--1679},
year={2018},
organization={IEEE}
}๐ Footwear 2018
(1) 1,000 object, 12 category footwear dataset, each object captured from 4 different poses.
@inproceedings{mahajan2018pose,
title={Pose Aware Fine-Grained Visual Classification Using Pose Experts},
author={Mahajan, Kushagra and Khurana, Tarasha and Chopra, Ayush and Gupta, Isha and Arora, Chetan and Rai, Atul},
booktitle={2018 25th IEEE International Conference on Image Processing (ICIP)},
pages={2381--2385},
year={2018},
organization={IEEE}
}๐ iMaterialist 2019
(1) 1M+ fashion images with 228 fine-grained attributes in total (multi-labeled).
(2) Category(105), Color(21), Gender(3), Material(34), Neckline(11), Print(28), Sleeve length(5), Design details(21).
[homepage] [pdf] [github] ๐ท
@inproceedings{guo2019imaterialist,
title={The imaterialist fashion attribute dataset},
author={Guo, Sheng and Huang, Weilin and Zhang, Xiao and Srikhanta, Prasanna and Cui, Yin and Li, Yuan and Adam, Hartwig and Scott, Matthew R and Belongie, Serge},
booktitle={Proceedings of the IEEE International Conference on Computer Vision Workshops},
pages={0--0},
year={2019}
}๐ Atlas 2019
(1) 186,150 images under clothing category with 3 levels and 52 leaf nodes in the taxonomy.
[homepage] [pdf] ๐ท
@article{umaashankar2019atlas,
title={Atlas: A Dataset and Benchmark for E-commerce Clothing Product Categorization},
author={Umaashankar, Venkatesh and Prakash, Aditi and others},
journal={arXiv preprint arXiv:1908.08984},
year={2019}
}๐ DeepShoe 2019
(1) 14,314 and 31,048 images from the street and online shop scenario in multiple viewpoints.
(2) Attributes including 22 Colors, 4 Toe shape, 7 Heel shape.
[homepage] [pdf]
@article{zhan2019deepshoe,
title={DeepShoe: An improved Multi-Task View-invariant CNN for street-to-shop shoe retrieval},
author={Zhan, Huijing and Shi, Boxin and Duan, Ling-Yu and Kot, Alex C},
journal={Computer Vision and Image Understanding},
volume={180},
pages={23--33},
year={2019},
publisher={Elsevier}
}๐ FindFashion 2019 combine DeepFashion & Street2Shop
(1) Customer-to-shop clothes retrieval dataset consists of 565,041 images and 382,230 pairs.
(2) Labeled 3 attributes (i.e., occlusions, views, and cropping).
(3) Divided the benchmark into 4 levels. i.e., Easy, Hard-Cropping, Hard-Occlusion, and Hard-View.
[homepage] [pdf]
@inproceedings{kuang2019fashion,
title={Fashion Retrieval via Graph Reasoning Networks on a Similarity Pyramid},
author={Kuang, Zhanghui and Gao, Yiming and Li, Guanbin and Luo, Ping and Chen, Yimin and Lin, Liang and Zhang, Wayne},
booktitle={Proceedings of the IEEE International Conference on Computer Vision},
pages={3066--3075},
year={2019}
}๐ GarmentSet 2020
(1) 9,636 images with collar part annotations and 8,616 images with shoulder and sleeve annotations.
(2) With annotation of landmarks of collars and sleeves on clean garment images.
[homepage] [pdf]
@inproceedings{chen2020tailorgan,
title={TailorGAN: Making User-Defined Fashion Designs},
author={Chen, Lele and Tian, Justin and Li, Guo and Wu, Cheng-Haw and King, Erh-Kan and Chen, Kuan-Ting and Hsieh, Shao-Hang and Xu, Chenliang},
booktitle={The IEEE Winter Conference on Applications of Computer Vision},
pages={3241--3250},
year={2020}
}For outfit generation, recommendation, evaluation, comnpatibility learning etc.
๐ WoW 2012
(1) 24,417 clothing images that are fully annotated.
(2) 7 multi-value clothing attributes and 10 occasion categories.
(3) 9,469 images with visible full-body. 8,421 images with only upper-body. 6,527 images with lower-body clothing.
(4) Color(11), Material(6), Print(6), Neckline(6), Sleeve length(3), Bottom length(3).
(5) Labeled with 10 common occasions.
[homepage] [pdf]
@inproceedings{liu2012hi,
title={Hi, magic closet, tell me what to wear!},
author={Liu, Si and Feng, Jiashi and Song, Zheng and Zhang, Tianzhu and Lu, Hanqing and Xu, Changsheng and Yan, Shuicheng},
booktitle={Proceedings of the 20th ACM international conference on Multimedia},
pages={619--628},
year={2012}
}๐ Stylatrix
[homepage] [[pdf]]
@article{sunstylatrix,
title={Stylatrix: an interactive model-based system for fashion exploration and outfit discovery},
author={Sun, Will J and Gajos, Krzysztof Z}
}๐ Edge2Garment 2016
[homepage] [pdf] [github] [datalink] ๐ท
@article{pix2pix2016,
title={Image-to-Image Translation with Conditional Adversarial Networks},
author={Isola, Phillip and Zhu, Jun-Yan and Zhou, Tinghui and Efros, Alexei A},
journal={arxiv},
year={2016}
}๐ FashionVC 2017
(1) 20,726 outfits with 14,871 tops and 13,663 bottoms.
(2) Visual image, categories and title description are collected.
[homepage] [pdf]
@inproceedings{song2017neurostylist,
title={Neurostylist: Neural compatibility modeling for clothing matching},
author={Song, Xuemeng and Feng, Fuli and Liu, Jinhuan and Li, Zekun and Nie, Liqiang and Ma, Jun},
booktitle={Proceedings of the 25th ACM international conference on Multimedia},
pages={753--761},
year={2017}
}๐ Fashion409K 2017
(1) 409,776 sets of clothing items from 644,192 unique items.
[homepage] [pdf] ๐ท
@InProceedings{Tangseng_2017_ICCV,
author = {Tangseng, Pongsate and Yamaguchi, Kota and Okatani, Takayuki},
title = {Recommending Outfits From Personal Closet},
booktitle = {The IEEE International Conference on Computer Vision (ICCV) Workshops},
month = {Oct},
year = {2017}
}๐ Polyvore 2017
(1) 21,889 outfits and 164,379 items.
(2) Keep the first 8 for simplicity, the average number of fashion items in an outfit is 6.5.
(3) To clean the text descriptions, words appearing fewer than 30 times are deleted, leading to a vocabulary of size 2,757.
[homepage] [pdf] [github] ๐ท
@inproceedings{han2017learning,
author = {Han, Xintong and Wu, Zuxuan and Jiang, Yu-Gang and Davis, Larry S},
title = {Learning Fashion Compatibility with Bidirectional LSTMs},
booktitle = {ACM Multimedia},
year = {2017},
}๐ AVA 2017
(1) Photo.net dataset contains 20,278 images with at least 10 score ratings per image.
(2) CUHK-PhotoQuality (CUHK-PQ) dataset t contains 17,690 images. All images are given a binary aesthetic label.
(3) Aesthetic Visual Analysis (AVA) dataset contains 250k images. Each image receives 78 โผ 549 votes of score ranging from 1 to 10.
[homepage] [pdf] ๐ท
@article{deng2017image,
author = {Deng, Yubin and Loy, Chen Change and Tang, Xiaoou},
title = {Image Aesthetic Assessment: An Experimental Survey},
journal={IEEE Signal Processing Magazine},
volume={34},
number={4},
pages={80--106},
year={2017},
publisher={IEEE}
}๐ UIUC 2018
(1) 68,306 outfits and 365,054 items.
(2) 19 max items, which has semantic category labeled.
@inproceedings{VasilevaECCV18FasionCompatibility,
Author = {Mariya I. Vasileva and Bryan A. Plummer and Krishna Dusad and Shreya Rajpal and Ranjitha Kumar and David Forsyth},
Title = {Learning Type-Aware Embeddings for Fashion Compatibility},
booktitle = {ECCV},
Year = {2018}
}๐ IQON 2018
(1) 164,837 items of clothing grouped in 21,889 outfits.
@article{nakamura2018outfit,
title={Outfit generation and style extraction via bidirectional lstm and autoencoder},
author={Nakamura, Takuma and Goto, Ryosuke},
journal={arXiv preprint arXiv:1807.03133},
year={2018}
}๐ Style4BodyShape 2018
(1) 3,150 female celebrities annotated with the corresponding types of body shapes.
(2) 349,298 images of 270 stylish celebrities annotated with the types of clothing items.
[homepage] [pdf]
@inproceedings{hidayati2018dress,
title={What dress fits me best? fashion recommendation on the clothing style for personal body shape},
author={Hidayati, Shintami Chusnul and Hsu, Cheng-Chun and Chang, Yu-Ting and Hua, Kai-Lung and Fu, Jianlong and Cheng, Wen-Huang},
booktitle={Proceedings of the 26th ACM international conference on Multimedia},
pages={438--446},
year={2018}
}๐ Lookastic 2019
(1) 30,790 fashionable outfits from the Lookastic.
(2) 124,665 matched pairs for men with 5,069 items.
(3) 158,755 matched pairs for women with 10,016 items.
(4) 65 Colors, 38 Materials, 40 Print, 253 fine-grained Categories, 11 Styles, and 114 Sub-categories.
[homepage] [pdf]
@inproceedings{yang2019interpretable,
title={Interpretable Fashion Matching with Rich Attributes},
author={Yang, Xun and He, Xiangnan and Wang, Xiang and Ma, Yunshan and Feng, Fuli and Wang, Meng and Chua, Tat-Seng},
booktitle={Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval},
pages={775--784},
year={2019}
}๐ POG 2019
(1) 1.01 million outfits, 583K fashion items, with context information.
(2) 0.28 billion user click actions from 3.57 million users.
[homepage] [pdf] [github] ๐ท
@inproceedings{chen2019pog,
title={POG: Personalized Outfit Generation for Fashion Recommendation at Alibaba iFashion},
author={Chen, Wen and Huang, Pipei and Xu, Jiaming and Guo, Xin and Guo, Cheng and Sun, Fei and Li, Chao and Pfadler, Andreas and Zhao, Huan and Zhao, Binqiang},
booktitle={Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery \& Data Mining},
pages={2662--2670},
year={2019}
}๐ Shop the Look 2019
(1) Based on Shop the Look (STL) task, and covert STL data into a format that can be used for our Complete the Look (CTL) task.
@inproceedings{kang2019complete,
title={Complete the Look: Scene-based Complementary Product Recommendation},
author={Kang, Wang-Cheng and Kim, Eric and Leskovec, Jure and Rosenberg, Charles and McAuley, Julian},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
pages={10532--10541},
year={2019}
}๐ ExpFashion 2019
(1) FashionVC+, consisting of both the visual and textual metadata of fashion items (i.e.,the tops, bottoms and shoes) on Polyvore.
(2) 20,726 outfits with 14,870 tops, 13,662 bottoms and 14,093 pairs of shoes, respectively.
(3) ExpFashion dataset consists of 893,991 outfits with 168,682 tops and 117,668 bottom.
[homepage] [pdf]
@article{liu2019neural,
title={Neural fashion experts: I know how to make the complementary clothing matching},
author={Liu, Jinhuan and Song, Xuemeng and Chen, Zhumin and Ma, Jun},
journal={Neurocomputing},
volume={359},
pages={249--263},
year={2019},
publisher={Elsevier}
}๐ ASOS outfits 2019
(1) 586,320 fashion outfits (images and textual descriptions) composed by ASOS stylists.
(2) Each containing between 2 and 5 items.
(3) 591,725 unique items representing 18 different womenswear product types and 22 different menswear product type.
[homepage] [pdf]
@article{bettaney2019fashion,
title={Fashion Outfit Generation for E-commerce},
author={Bettaney, Elaine M and Hardwick, Stephen R and Zisimopoulos, Odysseas and Chamberlain, Benjamin Paul},
journal={arXiv preprint arXiv:1904.00741},
year={2019}
}๐ Chuanda 2020
(1) 3,557 outfits covering 67 basic fashion styles.
(2) Labeled with 1,879 distinct fashion related attributes that belong to 5 types: Gender, Season, Style, Material, and Function.
[homepage] [pdf]
@article{liu2020imitation,
title={Imitation Learning for Fashion Style Based on Hierarchical Multimodal Representation},
author={Liu, Shizhu and Yang, Shanglin and Zhou, Hui},
journal={arXiv preprint arXiv:2004.06229},
year={2020}
}For 3D generation, VTON etc.
๐ MPI Dynamic FAUST | BUFF 2017
(1) 40,000 raw and aligned meshes.
[homepage] [pdf] ๐ท
@inproceedings{dfaust:CVPR:2017,
title = {Dynamic {FAUST}: {R}egistering Human Bodies in Motion},
author = {Bogo, Federica and Romero, Javier and Pons-Moll, Gerard and Black, Michael J.},
booktitle = {IEEE Conf. on Computer Vision and Pattern Recognition (CVPR)},
month = jul,
year = {2017},
month_numeric = {7}
}๐ FashionGAN 2017
(1) Extended the DeepFashion by collecting sentence descriptions for 79K images.
[homepage] [pdf] [github] ๐ท
@inproceedings{zhu2017be,
title={Be Your Own Prada: Fashion Synthesis with Structural Coherence},
author={Zhu, Shizhan and Fidler, Sanja and Urtasun, Raquel and Lin, Dahua and Chen, Change Loy},
booktitle={Proceedings of the IEEE Conference on International Conference on Computer Vision},
year={2017}
}๐ BeautyGAN 2018
(1) Facial makeup dataset consists of 3,834 female images.
[homepage] [pdf] ๐ท
@inproceedings{li2018beautygan,
title={Beautygan: Instance-level facial makeup transfer with deep generative adversarial network},
author={Li, Tingting and Qian, Ruihe and Dong, Chao and Liu, Si and Yan, Qiong and Zhu, Wenwu and Lin, Liang},
booktitle={Proceedings of the 26th ACM international conference on Multimedia},
pages={645--653},
year={2018}
}๐ VTON 2018
(1) Collected dataset from Zalando.
(2) 19,000 frontal-view woman and top2 image pairs (removed noisy images with no parsing results), yielding 16,253 pairs.
@inproceedings{han2017viton,
title = {VITON: An Image-based Virtual Try-on Network},
author = {Han, Xintong and Wu, Zuxuan and Wu, Zhe and Yu, Ruichi and Davis, Larry S},
booktitle = {CVPR},
year = {2018},
}๐ DeepWear 2018
(1) A specific brand clothes dataset.
[homepage] [pdf]
@inproceedings{kato2018deepwear,
title={DeepWear: a Case Study of Collaborative Design between Human and Artificial Intelligence},
author={Kato, Natsumi and Osone, Hiroyuki and Sato, Daitetsu and Muramatsu, Naoya and Ochiai, Yoichi},
booktitle={Proceedings of the Twelfth International Conference on Tangible, Embedded, and Embodied Interaction},
pages={529--536},
year={2018}
}๐ FashionGEN 2018
(1) 293,008 high-resolution fashion images paired with item descriptions provided by professional stylists.
(2) All fashion items are photographed from 1 to 6 different angles depending on the category of the item.
(3) 48 main categories, and 121 fine-grained sub-categories.
(4) Paired with paragraph-length descriptive captions sourced from experts (professional designers).
(5) Provide metadata such as stylist recommended matched items, the fashion season, designer and the brand.
(6) Provide the distribution of colors extracted from the text description.
@article{rostamzadeh2018fashion,
title={Fashion-gen: The generative fashion dataset and challenge},
author={Rostamzadeh, Negar and Hosseini, Seyedarian and Boquet, Thomas and Stokowiec, Wojciech and Zhang, Ying and Jauvin, Christian and Pal, Chris},
journal={arXiv preprint arXiv:1806.08317},
year={2018}
}๐ ZalandoGAN 2018
(1) Over 120,000 images of dresses that are downloaded from Zalandoโs website.
@article{yildirim2018disentangling,
title={Disentangling multiple conditional inputs in gans},
author={Yildirim, G{\"o}khan and Seward, Calvin and Bergmann, Urs},
journal={arXiv preprint arXiv:1806.07819},
year={2018}
}๐ RTW 2018
(1) Augment the dataset of 4,157 images by a factor 5 by jittering images with random scaling and translations.
(2) 7 categories: jackets, coats, shirts, tops, t-shirts, dresses and pullovers.
(3) 7 print: plain(uniform), plain(tiled?), striped, animal print(animal skin), dotted, graphic print(print and graphical pattern).
[homepage] [pdf]
@inproceedings{sbai2018design,
title={Design: Design inspiration from generative networks},
author={Sbai, Othman and Elhoseiny, Mohamed and Bordes, Antoine and LeCun, Yann and Couprie, Camille},
booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},
pages={0--0},
year={2018}
}๐ SMPL 2018
(1) Comprises a pair of jeans, a T-shirt and a sweater worn by 600 bodies in various poses.
@inproceedings{gundogdu19garnet,
title = {Garnet: A Two-stream Network for Fast and Accurate 3D Cloth Draping},
author = {Gundogdu, Erhan and Constantin, Victor and Seifoddini, Amrollah and Dang, Minh and Salzmann, Mathieu and Fua, Pascal},
booktitle = {{IEEE} International Conference on Computer Vision ({ICCV})},
month = {oct},
organization = {{IEEE}},
year = {2019},
}๐ MVP 2019
(1) Contains 35,687 person images and 13,524 clothes images. (2) Each person image in MPV has different poses. (3) The image is in the resolution of 256 ร 192. (4) 62,780 three-tuples of the same person in the same clothes but with different poses.
[homepage] [pdf] ๐ท
@inproceedings{dong2019towards,
title={Towards multi-pose guided virtual try-on network},
author={Dong, Haoye and Liang, Xiaodan and Shen, Xiaohui and Wang, Bochao and Lai, Hanjiang and Zhu, Jia and Hu, Zhiting and Yin, Jian},
booktitle={Proceedings of the IEEE International Conference on Computer Vision},
pages={9026--9035},
year={2019}
}๐ Fashion Takes 2019
(1) Over 18,000 images with meta-data including clothing category.
(2) Manual shape annotation indicating whether the personโs shape is above average or average.
(3) The data comprises 181 different users.
(4) Allowed to study ** the relationship between clothing categories and body shape**.
[homepage] [pdf]
@inproceedings{sattar2019fashion,
title={Fashion is taking shape: Understanding clothing preference based on body shape from online sources},
author={Sattar, Hosnieh and Pons-Moll, Gerard and Fritz, Mario},
booktitle={2019 IEEE Winter Conference on Applications of Computer Vision (WACV)},
pages={968--977},
year={2019},
organization={IEEE}
}๐ StyleGAN 2019
(1) Based on a proprietary image dataset with around 380k entries with high-resolution.
(2) An outfit is composed of a set of the maximum of 6 articles.
[homepage] [pdf]
@inproceedings{yildirim2019generating,
title={Generating High-Resolution Fashion Model Images Wearing Custom Outfits},
author={Yildirim, Gokhan and Jetchev, Nikolay and Vollgraf, Roland and Bergmann, Urs},
booktitle={Proceedings of the IEEE International Conference on Computer Vision Workshops},
pages={0--0},
year={2019}
}๐ Deep Fashion3D 2020
(1) 2,078 models reconstructed from real garments, which covers 10 different categories and 563 garment instances.
(2) Multi-view stereo, multi-view real images, 3D feature lines, 3D body pose.
@article{zhu2020deep,
title={Deep Fashion3D: A Dataset and Benchmark for 3D Garment Reconstruction from Single Images},
author={Zhu, Heming and Cao, Yu and Jin, Hang and Chen, Weikai and Du, Dong and Wang, Zhangye and Cui, Shuguang and Han, Xiaoguang},
journal={arXiv preprint arXiv:2003.12753},
year={2020}
}๐ Amazon Reviews 2015
(1) Based on the Amazon web store.
(2) Over 180 million relationships between a pool of almost 6 million objects.
@inproceedings{mcauley2015image,
title={Image-based recommendations on styles and substitutes},
author={McAuley, Julian and Targett, Christopher and Shi, Qinfeng and Van Den Hengel, Anton},
booktitle={Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval},
pages={43--52},
year={2015}
}๐ Tradesy 2016
(1) 19,823 users, 166,526 items, 410,186 feedback.
(2) Contains usersโ purchase histories and โthumbs-upโ.
[homepage] [pdf] ๐ท
@inproceedings{he2016vbpr,
title={VBPR: visual bayesian personalized ranking from implicit feedback},
author={He, Ruining and McAuley, Julian},
booktitle={Thirtieth AAAI Conference on Artificial Intelligence},
year={2016}
}๐ Fashion-MNIST 2017
(1) Comprising of 28 ร 28 grayscale images of 70,000 fashion products from 10 categories.
[homepage] [pdf] [github] ๐ท
@online{xiao2017/online,
author = {Han Xiao and Kashif Rasul and Roland Vollgraf},
title = {Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms},
date = {2017-08-28},
year = {2017},
eprintclass = {cs.LG},
eprinttype = {arXiv},
eprint = {cs.LG/1708.07747},
}๐ Shoe 2018
(1) Supports further research on the task of relative image captioning.
(2) Pairing 3,600 captions that were discriminative with additional dissimilar images.
[homepage] [pdf]
@inproceedings{guo2018dialog,
title={Dialog-based interactive image retrieval},
author={Guo, Xiaoxiao and Wu, Hui and Cheng, Yu and Rennie, Steven and Tesauro, Gerald and Feris, Rogerio},
booktitle={Advances in Neural Information Processing Systems},
pages={678--688},
year={2018}
}๐ Flickr30k 2018
(1) 300k posts with 5M comments.
(2) Each image is paired with user comment. The maximum number of comments is 427, average per image is 14.
[homepage] [pdf]
@inproceedings{lin2018netizen,
title={Netizen-style commenting on fashion photos: dataset and diversity measures},
author={Lin, Wen Hua and Chen, Kuan-Ting and Chiang, Hung Yueh and Hsu, Winston},
booktitle={Companion Proceedings of the The Web Conference 2018},
pages={395--402},
year={2018}
}๐ FCDB 2019
(1) 100 million Flickr images which focus on 21 global cities based on city perception.
(2) 25,707,690 clothing images for trend analysis.
[homepage] [pdf] [github] ๐ท
@InProceedings{Kataoka_2019_CVPR_Workshops,
author = {Kataoka, Hirokatsu and Satoh, Yutaka and Abe, Kaori and Minoguchi, Munetaka and Nakamura, Akio},
title = {Ten-Million-Order Human Database for World-Wide Fashion Culture Analysis},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2019}
}๐ Fashion IQ 2019
(1) Dataset for natural language based fashion image retrieval.
(2) Each image is crawled from Amazon.com and extracted corresponding product information, when available.
[homepage] [pdf] [github] [github] ๐ท
@article{guo2019fashion,
title={The Fashion IQ Dataset: Retrieving Images by Combining Side Information and Relative Natural Language Feedback},
author={Guo, Xiaoxiao and Wu, Hui and Gao, Yupeng and Rennie, Steven and Feris, Rogerio},
journal={arXiv preprint arXiv:1905.12794},
year={2019}
}๐ TFCD 2019
(1) A dataset containing data from real user sessions on a major European e-commerce fashion website.
[homepage] [pdf]
@article{bigon2019prediction,
title={Prediction is very hard, especially about conversion. Predicting user purchases from clickstream data in fashion e-commerce},
author={Bigon, Luca and Cassani, Giovanni and Greco, Ciro and Lacasa, Lucas and Pavoni, Mattia and Polonioli, Andrea and Tagliabue, Jacopo},
journal={arXiv preprint arXiv:1907.00400},
year={2019}
}- MPV (Multi-Pose Virtual try on) dataset [link]
- Clothing Attributes Dataset [link]
- Clothing Detection Dataset [link]
- Clothing Size Recommendation [link]
- Fashion Product Image [link]
- Dresses_Attribute_Sales Data Set [link]
- HICP-Clothing [link]
- Paper Doll Raw Dataset [link]
- Fashion Toolbox [link]
- https://data.world/datasets/fashion
- https://tianchi.aliyun.com/dataset/
Here I would like to thank Miss Po Yee(Boey), PANG, Miss Wai Lee(Selene), CHONG, for their hard work on collecting the datasource information.