pedestrain datasets

RAP, PETA, PA-100K

Most dataset and papers links

Most dataset

Simple introduction

  • Clothing Attributes Dataset: cloth style(7 categories), ensurely used. accuracy (about 55%)

  • PETA Dataset: low-resolution, multi-class(age 4 categories), download. accuracy(16,31,46,>61: 83.8, 78.8,76.4,89.0),19,000 images, about 0.82

  • Database of Human Attributes (HAT): high-resolution, age(multi-classes), having sent email

  • RAP Dataset: not high-resolution, 3-multi-classes(age 3 categories), waiting to send email.

  • PA-100K Dataset: binary

  • WIDER Attribute Dataset: multi-people

  • Parse27k Dataset: squence, oritation(multi-class), not so much related

  • CRP Dataset: squence, not human centre

  • Berkeley-Attributes of People: binary

  • Deepfashion dataset: cloth related, 289,222 - 50 classes cloth categories: 82%