OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Monoloco

Monoloco is a Python computer-vision library that uses 2D human pose keypoints to estimate 3D human locations from monocular or stereo images, with workflows for body orientation and social-distancing analysis.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Monoloco — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Monoloco when developing or researching a Python image-processing workflow that starts from human pose keypoints and needs monocular or stereo 3D localization. Review camera intrinsics when interpreting absolute distances.

Who it is for

  • Computer-vision developers building image-based human-localization workflows.
  • Researchers exploring monocular or stereo 3D pedestrian localization, orientation estimation, or social-distancing analysis.

Documented uses

  • Estimate people’s 3D locations from monocular images.
  • Estimate people’s 3D locations using stereo images.
  • Obtain 2D human keypoints through the documented OpenPifPaf mode.
  • Explore body-orientation and social-distancing workflows.
Before you choose

Requirements & limits

The README documents installation with pip. It identifies the software as AGPL-licensed and notes commercial licensing through the EPFL Technology Transfer Office.

  • The workflow relies on OpenPifPaf for 2D human pose detection before 3D localization.
  • The command interface distinguishes monocular, stereo, and keypoints-only modes.
  • Absolute distance estimates are affected by camera intrinsics; the README describes KITTI-specific parameters and default nuScenes camera assumptions for other cases.
  • The README says a GPU is optional but recommended for real-time performance.
  • The project documents JSON output and frontal or bird’s-eye visualizations.

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

Install Monoloco and inspect its command options

Install the library with pip and view the documented command-line options.

Before you start

  • Python 3.6, 3.7, or 3.8 is listed for the tested installation setup.
  • The README describes package installation with pip and virtual environments.
  • A GPU is optional; the README recommends one for real-time performance.

Steps

  1. Ensure there is no folder named monoloco in the current directory.

  2. Run the command below.

    Documented command
    pip3 install monoloco

  3. Run the command below to inspect the available subcommands.

    Documented command
    python3 -m monoloco.run --help

Expected outcome

The documented package installation is followed by the main command-line help.

Guide limitations

  • The README says installation was tested on OSX and Linux with the listed Python versions.
  • This setup guide does not run a prediction or verify a local installation.
Documentation 1 ↗
Source-based practical guide

Run a monocular prediction on an image

Use the documented prediction interface to request monocular 3D localization for an image.

Before you start

  • Monoloco installed in a Python environment.
  • An input image and an output directory.
  • The README example also supplies a ground-truth JSON file.

Steps

  1. Use the command below with the image path.

    Documented command
    python3 -m monoloco.run predict

  2. Select monocular mode with --mode mono, which the README identifies as the default.

  3. Set an output directory with -o and, if needed, set image scaling with --long-edge.

  4. Add --path_gt with a ground-truth JSON file when using the documented comparison example.

Expected outcome

The prediction command requests monocular human 3D localization; documented output options include visualizations and JSON.

Guide limitations

  • The README states that models are downloaded automatically.
  • Absolute distances depend on camera intrinsic parameters; the documented focal-length argument is --focal.
  • The README example includes a ground-truth file, but does not make that file a general requirement for prediction.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
18 October 2026
Original website title
vita-epfl/monoloco · Official homepage title · retained in its source language
How we review sourcesSuggest a correction