Hugging Face's logo

Spaces:
hugging-apps
/
webbrain-vl-2-450m-demo
Running on Zero

App Files Community
webbrain-vl-2-450m-demo / README.md
multimodalart's picture
multimodalart HF Staff
Upload folder using huggingface_hub
95ab6f4 verified
preview code
|
Raw
History Blame
1.41 kB
metadata 
title: WebBrain VL 2 450M
emoji: 🖥️
colorFrom: purple
colorTo: indigo
sdk: gradio
sdk_version: 6.26.0
app_file: app.py
short_description: Browser-GUI screenshot to structured observation text
python_version: '3.12'
startup_duration_timeout: 30m

WebBrain VL 2 450M — Browser Screenshot Understanding

WebBrain VL 2 450M is a compact vision-language model fine-tuned from LiquidAI/LFM2.5-VL-450M specifically for browser-GUI screenshot understanding. It converts a viewport screenshot into a terse, six-section observation consumed by a downstream web-automation planner.

What it does

Upload a browser screenshot and the model returns a structured description with:

  1. Page purpose — one-line summary
  2. Visible text — exact strings on buttons, links, headings, tabs
  3. Inputs — form fields with labels, placeholders, values, focus/disabled state
  4. State signals — loading spinners, toasts, modals, errors, CAPTCHAs, consent banners
  5. Blockers — anything preventing the next action
  6. Unknowns — things it can't read clearly

License

This model is distributed under the LFM Open License v1.0. Example screenshots are from the WebBrain test suite (CC-BY-4.0 / MIT).