OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Image & Video AI

Mivos

MiVOS is a computer-vision research project for interactive video object segmentation, combining interaction-to-mask processing, temporal propagation and difference-aware fusion. Its main repository includes an interactive GUI and tools for evaluating interactive segmentation on DAVIS.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Mivos — OpenViST project profile in Image & Video AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use the main MiVOS repository when you need its interactive GUI or DAVIS interactive evaluation workflow. For the README's DAVIS/YouTube semi-supervised workflow and propagation-module training, follow its direction to the separate Mask-Propagation repository.

Who it is for

  • Researchers studying interactive video object segmentation.
  • Developers building video annotation workflows that use human guidance to segment objects across frames.
  • Practitioners who want a GUI for interactive segmentation of a video or image sequence.

Documented uses

  • Guide object masks through a video using interactive input.
  • Run the provided GUI on a video file or a folder of images.
  • Evaluate interactive video object segmentation on DAVIS.
  • Explore a modular segmentation method combining interaction, propagation and difference-aware mask fusion.
Before you choose

Requirements & limits

The project is publicly available on GitHub under the MIT license.

  • The GUI requires PyQt5; the README provides separate instructions for starting it with a video or image folder.
  • The documented setup downloads pretrained models with download_model.py and requires gdown.
  • The project separates interactive segmentation from semi-supervised propagation and interaction-module training across three repositories.
  • The README identifies an STCN branch as a better and faster version, distinct from the main branch.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Start the MiVOS interactive GUI

Open the documented GUI workflow with a video or image folder as input.

Before you start

  • Python environment with PyTorch 1.7.1, torchvision 0.8.2 and OpenCV 4.2.0 listed as development versions.
  • Cython, progressbar, davis-interactive and PyQt5 for the GUI; the README also lists other dependencies and notes that common packages such as Pillow are expected.
  • A video file or a folder of images; the repository includes example/example.mp4.

Steps

  1. Install PyTorch and torchvision using the official PyTorch installation guide.

  2. Install the remaining listed packages as described in the repository README.

  3. Run the command below to download the required models.

    Documented command
    python download_model.py

  4. Start the command below or the command below; add --num_objects to label multiple objects.

    Documented command
    python interactive_gui.py --video
    Documented command
    python interactive_gui.py --images

Expected outcome

The interactive GUI opens with the selected video or image sequence for user-guided object segmentation.

Guide limitations

  • The README says its dependency list is not exhaustive and that other common Python packages may be expected.
  • The README directs semi-supervised DAVIS/YouTube VOS work to the separate Mask-Propagation repository.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
hkchengrex/MiVOS · Official homepage title · retained in its source language
How we review sourcesSuggest a correction