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

Epipolarpose

Epipolarpose is an open-source PyTorch implementation of a self-supervised method for 3D human pose estimation. During training, it estimates 2D poses from multi-view images and applies epipolar geometry to derive 3D pose and camera geometry for training a 3D pose estimator. The repository describes inference from a single RGB image and includes a demo notebook, training and validation scripts, data-preparation and m

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Epipolarpose — 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 it when investigating self-supervised 3D pose estimation from multi-view geometry, or when examining the repository’s documented research workflow and single-image demo.

Who it is for

  • Computer-vision researchers exploring self-supervised 3D human pose estimation.
  • Developers comparing multi-view geometry approaches or using the implementation as a research baseline.

Documented uses

  • Study a learning method that does not require 3D ground-truth poses or camera extrinsics.
  • Train or validate a 3D pose estimator using multi-view image data.
  • Inspect a sample 3D pose prediction from an RGB image through the demo notebook.
Before you choose

Requirements & limits

The repository is public and identified as open source. Human3.6M download permission requires an account with the dataset provider.

  • The method uses multi-view images during training but is described as accepting a single RGB image at test time.
  • The README says the learning method does not need 3D ground-truth data or camera extrinsics.
  • Training and testing require an NVIDIA GPU according to the repository requirements.
  • The documented setup uses Python 3.7.1 on Ubuntu 16.04 and PyTorch 1.0 or newer.
  • Human3.6M data is needed for training or testing with that dataset; MPII data is described for optional pretraining.

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

Prepare the repository for its demo notebook

Set up the documented environment and workspace to reach the demo notebook stage.

Before you start

  • Python 3.7.1 on Ubuntu 16.04 is the environment described by the repository.
  • PyTorch 1.0 or newer.
  • An NVIDIA GPU is required for training and testing.
  • Repository dependencies, installed through pip or the provided Conda environment file.

Steps

  1. Clone the EpipolarPose repository.

  2. Install dependencies with the command below or create the environment from environment.yml.

    Documented command
    pip install -r requirements.txt

  3. Download the annotation archive linked in the README into the repository and unzip it.

  4. Create output and models directories.

  5. Open demo.ipynb, which the README identifies as the simple demo.

Expected outcome

The repository and its documented annotation files and workspace directories are prepared to proceed to the demo notebook.

Guide limitations

  • The README does not provide the notebook's execution steps here.
  • Human3.6M data is required for training or testing with that dataset; download permission requires an account with its provider.
Documentation 1 ↗
Evidence and transparency

Sources & review

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