对于数据集有学习科研等需求的,请在 AIUAI-Dataset - DeepFashion 服装数据集 中联系.
下载文件包括:
50 category_name category_type Anorak 1 带风帽的厚茄克;防水布;滑雪衫 Blazer 1 运动夹克,运动上衣 Blouse 1 短上衣;女衬衫;宽松的上衣;工装 Bomber 1 Bomber Jacket 飞行员夹克 Button-Down 1 (衬衫)领尖有纽扣的,纽扣领的 Cardigan 1 开襟羊毛衫 Flannel 1 法兰绒衣服;法兰绒,绒布;毛巾; Halter 1 吊带 Henley 1 亨利 Hoodie 1 连帽衫;带帽夹克; Jacket 1 短上衣,夹克 Jersey 1 针织 Parka 1 风雪大衣;派克大衣 Peacoat 1 水手穿的厚呢短大衣 Poncho 1 斗篷 Sweater 1 毛衣,运动衫 Tank 1 tank top 背心装 Tee 1 T恤;短袖圆领运动衫(等于T-shirt) Top 1 上衣 Turtleneck 1 高领绒衣;高翻领,圆翻领 Capris 2 女用紧身裤 Chinos 2 斜纹棉布裤 Culottes 2 女裙裤 Cutoffs 2 拼接款 Gauchos 2 南美牛仔 Jeans 2 牛仔裤;粗斜纹棉布裤 Jeggings 2 牛仔样式打底紧身裤;是jeans(牛仔裤)和leggings(打底紧身裤)两个词的合成词; Jodhpurs 2 骑马裤,短马靴 Joggers 2 慢跑裤 Leggings 2 (女式)紧身裤 Sarong 2 马来群岛土人所穿的围裙,布裙 Shorts 2 短裤 Skirt 2 裙子;边缘;(连衣裙、外衣等的)下摆 Sweatpants 2 运动裤 Sweatshorts 2 Trunks 2 (男式)游泳裤 Caftan 3 有腰带的长袖衣服 Cape 3 披肩;斗篷 Coat 3 上衣,外套 Coverup 3 Dress 3 连衣裙 Jumpsuit 3 连衣裤,(尤指女式)连衣裤 Kaftan 3 土耳其式长衫 Kimono 3 (日本的)和服;和服式女晨衣 Nightdress 3 (妇女或孩子穿的)睡衣,睡袍 Onesie 3 连身衣 Robe 3 长袍;罩袍; 浴袍;睡袍 Romper 3 背心连裤子的衣服 Shirtdress 3 (上身像衬衫的)衬衣式连衣裙 Sundress 3 太阳裙,背心裙
实际上应该只有 46 类服装类别:
[ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 39, 40, 41, 42, 43, 44, 46, 47, 48]
每类服装有的图片数:
[160, 7495, 24557, 309, 330, 13311, 324, 17, 716, 4048, 10467, 748, 676, 97, 791, 13123, 15429, 36887, 10078, 146, 77, 527, 486, 1669, 49, 7076, 594, 45, 4416, 5013, 32, 19666, 14773, 3048, 1106, 386, 54, 2120, 17, 72158, 6153, 126, 2294, 70, 150, 7408]
可视化:
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
datas = open('list_category_img.txt').readlines()[2:]
print 'Num of DeepFashion Category Images: ', len(datas)
ann_labels = [eval(data.split(' ')[-1].strip()) for data in datas]
category_labels = np.unique(ann_labels)
print 'Num of DeepFashion Category: ', len(category_labels)
label_counts = [ann_labels.count(label_temp) for label_temp in category_labels]
print 'Num of DeepFashion Each Category Images: ', label_counts
plt.figure(figsize = (12,6))
sns.barplot(category_labels, label_counts, alpha = 0.9)
plt.xticks(rotation = 'vertical')
plt.xlabel('Image Labels', fontsize =12)
plt.ylabel('Counts', fontsize = 12)
plt.show()
下载文件包括:
[1] - DeepFashion: Powering Robust Clothes Recognition and Retrieval with Rich Annotations
[2] - Fashion Landmark Detection in the Wild
[3] - 论文阅读理解 - DeepFashion: Powering Robust Clothes Recognition and Retrieval with Rich Annotations