id int64 0 1.05k | image imagewidth (px) 1.08k 1.08k | question stringlengths 10 98 | answers listlengths 1 1 | ui_element_bbox listlengths 0 4 | hierarchy_xml stringlengths 4.74k 1.25M | resource_id stringlengths 0 81 | file_name stringlengths 22 33 |
|---|---|---|---|---|---|---|---|
0 | Do que se trata o texto em negrito na tela? | [
"Política de privacidade e uso 99"
] | [
0.051899999380111694,
0.3456999957561493,
0.9480999708175659,
0.4187999963760376
] | <?xml version='1.0' encoding='UTF-8' standalone='yes' ?><hierarchy rotation="0"><node index="0" text="" resource-id="" class="android.widget.FrameLayout" package="com.taxis99" content-desc="" checkable="false" checked="false" clickable="false" enabled="true" focusable="false" focused="false" scrollable="false" long-cli... | com.taxis99:id/title | 99_20251213_152904.png | |
1 | Como se pode aceitar as políticas de uso? | [
"Clicando no botão Concordo"
] | [
0.051899999380111694,
0.7791000008583069,
0.9480999708175659,
0.8414999842643738
] | <?xml version='1.0' encoding='UTF-8' standalone='yes' ?><hierarchy rotation="0"><node index="0" text="" resource-id="" class="android.widget.FrameLayout" package="com.taxis99" content-desc="" checkable="false" checked="false" clickable="false" enabled="true" focusable="false" focused="false" scrollable="false" long-cli... | com.taxis99:id/okButton | 99_20251213_152904.png | |
2 | Como sair? | [
"Clicando no botão Sair"
] | [
0.051899999380111694,
0.8560000061988831,
0.9480999708175659,
0.91839998960495
] | <?xml version='1.0' encoding='UTF-8' standalone='yes' ?><hierarchy rotation="0"><node index="0" text="" resource-id="" class="android.widget.FrameLayout" package="com.taxis99" content-desc="" checkable="false" checked="false" clickable="false" enabled="true" focusable="false" focused="false" scrollable="false" long-cli... | com.taxis99:id/cancelButton | 99_20251213_152904.png | |
3 | Os botões principais tem preenchimento com cor? | [
"Sim"
] | [] | <?xml version='1.0' encoding='UTF-8' standalone='yes' ?><hierarchy rotation="0"><node index="0" text="" resource-id="" class="android.widget.FrameLayout" package="com.taxis99" content-desc="" checkable="false" checked="false" clickable="false" enabled="true" focusable="false" focused="false" scrollable="false" long-cli... | contextual_question | 99_20251213_152904.png | |
4 | Qual a loja com entrega mais rápida? | [
"Açaí Flash Jaap"
] | [
0.051899999380111694,
0.486299991607666,
0.41670000553131104,
0.5090000033378601
] | <?xml version='1.0' encoding='UTF-8' standalone='yes' ?><hierarchy rotation="0"><node index="0" text="" resource-id="" class="android.widget.FrameLayout" package="com.taxis99" content-desc="" checkable="false" checked="false" clickable="false" enabled="true" focusable="false" focused="false" scrollable="false" long-cli... | com.taxis99:id/home_recommendation_food_shop_name | 99_20251213_153121.png | |
5 | Quanto tempo a loja mais rápida demora para entregar? | [
"15-30 min"
] | [
0.051899999380111694,
0.5174999833106995,
0.1906999945640564,
0.5371999740600586
] | <?xml version='1.0' encoding='UTF-8' standalone='yes' ?><hierarchy rotation="0"><node index="0" text="" resource-id="" class="android.widget.FrameLayout" package="com.taxis99" content-desc="" checkable="false" checked="false" clickable="false" enabled="true" focusable="false" focused="false" scrollable="false" long-cli... | com.taxis99:id/home_recommendation_food_delivery_time | 99_20251213_153121.png | |
6 | Qual o valor da taxa de entrega? | [
"Grátis"
] | [
0.2295999974012375,
0.5174999833106995,
0.3138999938964844,
0.5371999740600586
] | "<?xml version='1.0' encoding='UTF-8' standalone='yes' ?><hierarchy rotation=\"0\"><node index=\"0\"(...TRUNCATED) | com.taxis99:id/home_recommendation_food_delivery_fee | 99_20251213_153121.png | |
7 | Qual a finalidade desse app? | [
"Realizar viagens, fornecer comida, fazer entregas"
] | [] | "<?xml version='1.0' encoding='UTF-8' standalone='yes' ?><hierarchy rotation=\"0\"><node index=\"0\"(...TRUNCATED) | contextual_question | 99_20251213_153121.png | |
8 | A interface está usando tema claro ou escuro? | [
"Claro"
] | [] | "<?xml version='1.0' encoding='UTF-8' standalone='yes' ?><hierarchy rotation=\"0\"><node index=\"0\"(...TRUNCATED) | contextual_question | 99_20251213_153121.png | |
9 | Qual a função do botão roxo? | [
"sem_resposta"
] | [] | "<?xml version='1.0' encoding='UTF-8' standalone='yes' ?><hierarchy rotation=\"0\"><node index=\"0\"(...TRUNCATED) | sem_resposta | 99_20251213_153121.png |
- 🌟 Context and Inspiration
- 🎯 Research Objective
- 📊 Dataset Statistics
- 🌟 Contexto e Inspiração
- 🇧🇷 Diferencial: Aplicativos Brasileiros e Localização
- 🎯 Objetivo da Pesquisa
- 📊 Estatísticas do Dataset
- Esta escala reduzida é intencional, priorizando a revisão humana meticulosa.
- license: cc-by-nc-4.0
task_categories:
- visual-question-answering
language:
- pt
pretty_name: Screen QA Portuguese
citation: |
@misc{brito2026screenqapt,
author = {Brito, Adriana Camargo de},
title = {Screen QA Portuguese: A Multimodal Dataset for Brazilian Mobile Interfaces},
year = {2026},
publisher = {Hugging Face},
journal = {Hugging Face Repositories},
howpublished = {\url{https://huggingface.co/datasets/britoadriana/screen_qa_portuguese}}
}
🌟 Context and Inspiration
This dataset is a direct contribution to the field of ScreenQA (Screen Question Answering). It was inspired by the original ScreenQA dataset by Google Research, which utilizes the Rico database to provide visual hierarchy annotations and screenshots. 🇧🇷 Differentiation: Brazilian Apps and Localization While the original dataset focuses predominantly on English-language interfaces and global contexts, this Final Capstone Project focuses on a Brazilian and Portuguese version. Content: Selection of Brazilian mobile application interfaces. Language: Questions and answers formulated entirely in Portuguese by a human, respecting technical terms and common expressions specific to mobile interfaces in the country.
🎯 Research Objective
The primary goal of this dataset is to serve as a benchmark for testing and evaluating Vision Language Models (VLMs) within the specific domain of Brazilian or Portuguese-language interfaces. The research aims to understand how effectively these multimodal models can: Interpret visual elements (icons, buttons, text fields). Understand spatial layout (bounding boxes) within mobile interfaces. Answer complex questions regarding navigation flow and content in native Portuguese.
📊 Dataset Statistics
In this initial version, Screen QA Portuguese features a compact and highly curated structure, prioritizing the quality of annotations: Total Images: 302 unique screenshots. Total Questions: 1,050 question-answer pairs (averaging ~3.4 questions per image). This smaller scale is intentional, focusing on meticulous human review.
🌟 Contexto e Inspiração
Este dataset é uma contribuição direta para a área de ScreenQA (Perguntas e Respostas sobre Telas). Ele foi inspirado no dataset original ScreenQA do Google Research, que utiliza a base de dados Rico para fornecer anotações de hierarquia visual e capturas de tela.
🇧🇷 Diferencial: Aplicativos Brasileiros e Localização
Enquanto o dataset original foca predominantemente em interfaces em inglês e contextos globais, este projeto de Trabalho Final de Curso foca na versão brasileira e em português.
- Conteúdo: Seleção de interfaces de aplicativos brasileiros.
- Idioma: Perguntas e respostas formuladas inteiramente em português, por um ser humano, respeitando termos técnicos e expressões comuns de interfaces móveis no país.
🎯 Objetivo da Pesquisa
O objetivo principal deste dataset é servir como um benchmark para testar e avaliar Vision Language Models (VLMs) no domínio específico de interfaces brasileiras ou em português.
A pesquisa busca entender quão bem esses modelos multimodais conseguem:
- Interpretar elementos visuais (ícones, botões, campos de texto).
- Compreender a disposição espacial (bounding boxes) em interfaces mobile.
- Responder a perguntas complexas sobre o fluxo de navegação e conteúdo em português nativo.
📊 Estatísticas do Dataset
Nesta versão inicial, o Screen QA Portuguese apresenta uma estrutura compacta e altamente curada, focada na qualidade das anotações:
- Total de Imagens: 302 capturas de tela exclusivas.
- Total de Perguntas: 1050 pares de pergunta-resposta (média de ~3.4 perguntas por imagem).
Esta escala reduzida é intencional, priorizando a revisão humana meticulosa.
license: cc-by-nc-4.0 task_categories: - visual-question-answering language: - pt pretty_name: Screen QA Portuguese citation: | @misc{brito2026screenqapt, author = {Brito, Adriana Camargo de}, title = {Screen QA Portuguese: A Multimodal Dataset for Brazilian Mobile Interfaces}, year = {2026}, publisher = {Hugging Face}, journal = {Hugging Face Repositories}, howpublished = {\url{https://huggingface.co/datasets/britoadriana/screen_qa_portuguese}} }
- Downloads last month
- 31