WeVisOne-VL-2B

WeVisOne-VL-2B

Tries Tencent's WeVisOne-VL-2B vision-language model on images, from finding and counting objects and reading text to cutting out subjects. For developers and researchers who work with image understanding.

Made by hugging-apps on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: WeVisOne-VL-2B is a Gradio demo of tencent/WeVisOne-VL-2B, one vision-language model built on Qwen3-VL. The model handles both object-level and pixel-level work on images, so one page covers detection, text reading and cut-outs. It runs on ZeroGPU hardware, so you can try it in a browser.

Who it suits: It suits developers and researchers who want to see what a single vision-language model can do across many image tasks, without installing anything.

Industry | Who uses it | How they use it | Example
Computer vision | Researchers | Test one model on many perception tasks | Run grounding and counting on the same picture
Software | Developers | Check whether the model fits an image feature | Try OCR on a photo that contains text
Design and media | Editors | Cut a subject out of a picture | Use alpha matting to separate a person from the background

If you only need plain image captions, you can skip it.

Where to use it:

• A web browser, on the Space's page on Hugging Face
• A local copy of the files, run on Python 3.12 under Windows, macOS or Linux

Use case | Who | What they get back
Find and count objects | Researchers | Boxes or counts for what is in an image
Read text in images | Developers | The text found in the picture
Cut out a subject | Editors | A mask or matte of the subject

How to set it up:

• A web browser, which is all the online page needs
• Python 3.12 and the packages in requirements.txt, for a local copy
• The system packages in packages.txt, installed with apt-get

1. Open the Space's page on Hugging Face to use it online. Nothing needs installing.
2. For a local copy, install the Hugging Face hub package and download the Space's files.
3. Move into the downloaded folder.
4. Install Gradio 6.30.0, then the packages from requirements.txt.
5. Start the app with python app.py.

pip install -U huggingface_hub
hf download hugging-apps/wevisone-vl-2b --repo-type space --local-dir wevisone-vl-2b
cd wevisone-vl-2b
pip install gradio==6.30.0
pip install -r requirements.txt
python app.py

How to use it: Open the page, or the one your local copy starts, load an image and pick the kind of perception task you want. The model returns the result for that image, drawn by the page's rendering helper.

Tool | What it does
Grounding | Finds the part of an image that a phrase refers to
Dense detection | Finds many objects in one image
Counting | Counts the objects of a kind
Pointing | Marks points on objects
OCR | Reads the text in an image
Grounded captioning | Describes an image and ties the words to places in it
Referring segmentation | Outlines the pixels of whatever a phrase refers to
Alpha matting | Separates a subject from its background with soft edges

Example: Load a street photo and ask for counting. The page returns how many of the chosen object it finds and marks them on the picture.

For newcomers and experts: Open the Space page and try the OCR task on a picture with a sign or a label. It is the easiest one to judge by eye.

The model is one unified vision-language model on Qwen3-VL, so object-level tasks like detection and pointing and pixel-level tasks like segmentation and matting share one set of weights. The wevisone_vl inference package, the rendering helper and the examples come from the model's official repository.

You load an image and pick a task
The wevisone_vl package runs the model
The model returns boxes, points, text or masks
wv_render.py draws the result on the image

Good to know: The tasks are grouped as object-level and pixel-level perception, and the page does not say which image sizes it accepts.

• License: Apache 2.0
• Cost to run: none named, and the online copy runs on ZeroGPU hardware
• Data it reaches: the images you load and the tencent/WeVisOne-VL-2B model

On Hugging Face, 13 November 2026

• Likes: 8
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/hugging-apps/wevisone-vl-2b
• Files: https://huggingface.co/spaces/hugging-apps/wevisone-vl-2b/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.