---
title: VIEW2SPACE-4B
emoji: 👁
colorFrom: pink
colorTo: purple
sdk: gradio
sdk_version: 6.19.0
app_file: app.py
short_description: Multi-view visual reasoning VLM based on Qwen3-VL 4B
python_version: "3.12"
startup_duration_timeout: 30m
---

# VIEW2SPACE-4B Demo

A Gradio demo for [Pokerme/view2space_4b](https://huggingface.co/Pokerme/view2space_4b),
a 4.4B-parameter vision-language model based on Qwen3-VL that performs multi-view
visual reasoning from sparse observations.

## Usage

Upload one or more images and type a question. The model integrates information
across all provided views to produce an answer. Examples include counting objects,
describing scenes, detecting regions, and answering multiple-choice spatial
reasoning questions.

## Reference

- **Paper**: [VIEW2SPACE: Studying Multi-View Visual Reasoning from Sparse Observations](http://arxiv.org/abs/2603.16506) (ECCV 2026)
- **Code**: [github.com/pokerme7777/VIEW2SPACE](https://github.com/pokerme7777/VIEW2SPACE)