Dataset: Faces extracted from Time Magazine 1923-2014

The data presented here consists of three parts: Dataset 1: In this set, we extract 327,322 faces from our entire collection of 3389 issues, and automatically classified each face as male or female. We present this data as a single table with columns identifying the date, issue, page number, the coordinates identifying the position of the face on the page, and classification (male or female). The coordinates identifying the position of the face on the page are based on the size and resolution of the pages found in the “Time Vault”. Dataset 2: Dataset 2 consists of 8,789 classified faces from 100 selected issues. Human labor was used to identify and extract 3,299 face images from 39 issues, which were later classified by another set of workers. This selection of 39 issues contains one issue per decade spanned by the archive plus one issue per year between 1961 and 1991, and the extracted face images were used to train the face extraction algorithm. The remaining 5,490 faces from 61 issues were extracted via machine learning before being classified by human coders. These 61 issues were chosen to complement the first selection of 39 issues: one issue per year for all years in the archive excluding those between 1961 and 1991. Thus, Dataset 2 contains fully-labelled faces from at least one issue per year. Dataset 3: In the interest of transparency, Dataset 3 consists of the raw data collected to create Dataset 2, and consists of 2 tables. Before explaining these tables we first briefly describe our data collection and verification procedures, which have been fully described elsewhere. A custom AMT interface was used to enable human labors to classify faces according the categories in Table 4. Each worker was given a randomly-selected batch of 25 pages, each with a clearly highlighted face to be categorized, of which three pages were verification pages with known features, which were used for quality control. Each face was labeled by two distinct human coders, determined at random so that the paring of coders varied with the image. A proficiency rating was calculated for each coder by considering all images they annotated and computing the average number of labels that matched those identified by the image’s other coder. The tables in Dataset 2 were created by resolving inconsistencies between the two image coders by selecting the labels from the coder with the highest proficiency rating. Prior to calculating the proficiency score, all faces that were tagged as having ‘Poor’ or ‘Error’ image quality by either of the two coders were eliminated. Due to technical bugs when the AMT interface was first implemented, a small number of images were only labeled once; these were also eliminated from Datasets 2 and 3. In Dataset 3, we present the raw annotations for each coder that tagged each face, along with demographic data for each coder. Dataset 3 consists of two tables: the raw data from each of the two sets of coders, and the demographic information for each of the coders.