Dataset Viewer
Auto-converted to Parquet Duplicate
image
imagewidth (px)
2.84k
2.84k
task_type
stringclasses
5 values
question
stringclasses
6 values
answer
stringclasses
6 values
source_url
stringclasses
1 value
image_type
stringclasses
1 value
domain
stringclasses
1 value
caption
Apibūdink, kas matoma šiame vaizde.
Ekrane matomas „Sodros“ svetainės puslapis gyventojams. Viršuje pateiktas navigacijos meniu su skiltimis „Pensijos“, „Išmokos“, „Įmokos“, „Formos ir šablonai“ ir „Skaičiuoklės“, taip pat matomas paieškos mygtukas bei mėlynas mygtukas „Prisijunkite“. Pagrindiniame bloke rodoma informacija apie valstybės paslaugas šeimom...
https://www.sodra.lt/
web_ui_screenshot
public_services
site_identification
Kokios institucijos svetainė matoma ekrane?
Ekrane matoma „Sodros“ svetainė.
https://www.sodra.lt/
web_ui_screenshot
public_services
navigation
Ką vartotojas turėtų paspausti norėdamas prisijungti?
Vartotojas turėtų paspausti mygtuką „Prisijunkite“ viršutiniame dešiniajame kampe.
https://www.sodra.lt/
web_ui_screenshot
public_services
navigation
Kokios pagrindinės paslaugų kategorijos matomos apatinėje ekrano dalyje?
Apatinėje ekrano dalyje matomos kategorijos „Pensijos“, „Išmokos“ ir „Įmokos“.
https://www.sodra.lt/
web_ui_screenshot
public_services
ui_understanding
Koks pagrindinis informacinis blokas matomas puslapio centre?
Puslapio centre matomas blokas „Informacija apie valstybės paslaugas šeimoms“ su mygtuku „Noriu sužinoti“.
https://www.sodra.lt/
web_ui_screenshot
public_services
action
Ką vartotojas turėtų paspausti norėdamas sužinoti daugiau apie valstybės paslaugas šeimoms?
Vartotojas turėtų paspausti mygtuką „Noriu sužinoti“.
https://www.sodra.lt/
web_ui_screenshot
public_services

LT UI VLM

LT UI VLM is a small educational dataset for adapting small vision-language models to Lithuanian user interface and visual instruction understanding.

The dataset contains screenshots of public Lithuanian websites and interfaces with Lithuanian caption and question-answer annotations. It is prepared for a LoRA fine-tuning experiment with small open-weight vision-language models.

Dataset Structure

lt-ui-vlm/
  README.md
  train/
    images/
      example_001.jpg
    metadata.jsonl
  validation/

Each row in metadata.jsonl is one training example. The same image can appear in multiple rows with different questions.

Minimal row format:

{"file_name":"images/example_001.jpg","question":"Ką vartotojas turėtų paspausti norėdamas prisijungti?","answer":"Vartotojas turėtų paspausti prisijungimo mygtuką viršutiniame dešiniajame kampe."}

Annotation Guidelines

For each image, the recommended annotation pattern is:

  • 1 caption-style row describing the main content of the image.
  • 2-4 QA rows about navigation, visible text, UI elements, or the expected user action.

The annotations are written in Lithuanian because the dataset is intended for Lithuanian VLM adaptation.

Intended Uses

  • Lithuanian UI screenshot captioning.
  • Visual question answering for Lithuanian web interfaces.
  • User action prediction from Lithuanian UI screenshots.
  • Educational LoRA fine-tuning experiments with small VLMs.

Limitations

  • This is a small educational dataset and is not representative of all Lithuanian user interfaces.
  • The dataset is intended for experimentation, not production use.
  • Only public pages or empty forms should be included.
  • Private, sensitive, or personally identifiable information should be avoided.
Downloads last month
29