Mobile Interfaces Dataset

User Interfaces Dataset for the Automated Assessment of Visual Aesthetics of Mobile User Interfaces with Deep Learning

On this page we provide the 3139 screenshots present in the dataset used during the development of the Appsthetics neural network model for automatic assessment of Android user interfaces.

Data Aquisition

This dataset is composed of images of screenshots of Android apps developed with App Inventor that are available in the App Inventor Gallery and Android interfaces of the RICO dataset.

The interfaces were selected manually to avoid duplication of images and to assure ethical aspects.

Data Formats

Data organization for classification:

Images in JPG (.jpg) format with 239 × 425 pixels, grouped in 4 groups (beautiful, more or less,  ugly, undecided) with respect to their visual aesthetic based on the results of a human rating process:

  • Several online questionnaires, each with 150 random images (without repetition) and 3 voting options (beautiful, more or less ugly), were generated via Google Forms.
  • Each questionnaire was answered by three different volunteers.
  • Images were grouped based on the agreement of at least 2 of the 3 votes. If there was no agreement on the visual aesthetics, the image was included in the ‘undecided’ category.

Data organization for regression:

JPG (.jpg) images with 239 × 408 pixels, all grouped into a single set, together with a CSV file (.csv) containing, per line, the image name, a comma, the image score, a comma, and a boolean value indicating whether the image is part of the test suite or not.

The visual aesthetics scores were calculated based on the same survey mentioned above, assigning numerical values to the response alternatives: beautiful = 1.0, more or less  = 0.5, ugly = 0.0. After conversion, an average was calculated from the total obtained. In this case, only images with total disagreement with the votes were disregarded.


Files available on AppInventor Gallery are licensed under CreativeCommons 4.0

The RICO dataset does not provide any information on licensing.

Citing this Dataset

  author = {Heiderscheidt Martins, O. and Gresse von Wangenheim, C. and von Wangenheim, A.},
  title = {User Interfaces Dataset for the Automated Assessment of Visual Aesthetics of Mobile User Interfaces with Deep Learning},
  year = {2019},
  publisher = {GQS/INCoD/UFSC},
  journal = {INCoD Datasets Repository},
  howpublished = {\url{}}

Keywords: Aesthetics, Mobile application, Android, Deep learning, Visual design


Sobre o Autor

possui graduação em Ciências da Computação pela Universidade Federal de Santa Catarina (1989) e Doutorado Acadêmico (Dr. rer.nat.) em Ciências da Computação pela Universidade de Kaiserslautern (1996). Atualmente é professor Titular da Universidade Federal de Santa Catarina, onde é professor do Programa de Pós-graduação em Ciência da Computação e dos cursos de graduação em Ciências da Computação e Sistemas de Informação. Tem experiência nas áreas de Informática em Saúde, Processamento e Análise de Imagens e Engenharia Biomédica, com ênfase em Telemedicina, Telerradiologia, Sistemas de Auxílio ao Diagnóstico por Imagem e Processamento de Imagens Médicas, com foco nos seguintes temas: analise inteligente de imagens, DICOM, CBIR, informática médica, visão computacional e PACS. Coordena o Instituto Nacional de Ciência e Tecnologia para Convergência Digital - INCoD. Foi o criador e primeiro Coordenador do Núcleo de Telessaúde de Santa Catarina no âmbito do Programa Telessaúde Brasil do Ministério da Saúde e da OPAS - Organização Pan-Americana de Saúde e criador do Núcleo Santa Catarina da RUTE - Rede Universitária de Telemedicina.