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

Openemma

Openemma is an open-source Python implementation of Waymo’s EMMA multimodal model approach for end-to-end autonomous-driving motion planning. It combines text and front-view camera inputs with supported vision-language models to predict future ego-vehicle waypoints and produce decision rationales.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Openemma — 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 Openemma to explore or run a dataset-based autonomous-driving motion-planning workflow and inspect its waypoint, rationale and visualization outputs.

Who it is for

  • Researchers studying multimodal motion planning for autonomous vehicles
  • Developers exploring end-to-end autonomous-driving model implementations

Documented uses

  • Generate predicted future vehicle waypoints from text and front-view camera inputs
  • Inspect textual rationales about scene context, critical objects and behavior decisions
  • Visualize predicted trajectories and detected critical objects in annotated images and compiled video
Before you choose

Requirements & limits

The repository identifies its code license as Apache 2.0 and documents installation through PyPI or from the repository.

  • The documented setup uses Python 3.8 in Conda.
  • The repository installation route specifies CUDA Toolkit and project dependencies.
  • The README lists GPT-4o, LLaVA-1.6-Mistral-7B, Llama-3.2-11B-Vision-Instruct and Qwen2-VL-7B-Instruct as supported models.
  • GPT-4 use requires an OpenAI API key; nuScenes data is needed for the documented run.
  • The workflow produces research outputs and visualizations rather than a documented vehicle deployment.

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

Set up Openemma and prepare a model run

Install Openemma, configure a documented model option and run it with nuScenes data to generate planning outputs.

Before you start

  • Python 3.8 and Conda
  • CUDA Toolkit and project requirements for the repository installation route
  • An extracted nuScenes dataset
  • An OpenAI API key if using GPT-4

Steps

  1. Create and activate a Conda environment with Python 3.8.

  2. Install Openemma using the command below, or use the documented repository route and install its requirements.

    Documented command
    pip install openemma

  3. If using GPT-4, provide an OpenAI API key as an environment variable or in the code where prompted.

  4. Run the documented PyPI command with --model-path, --dataroot, --version and --method openemma, selecting a supported model option.

Expected outcome

The documented outputs include future ego-vehicle waypoints, decision rationales, annotated images and a compiled video.

Guide limitations

  • The documented run requires nuScenes data.
  • The README specifies CUDA Toolkit for the repository installation route.
  • GPT-4 reasoning requires an OpenAI API key.
  • The documented outputs describe model predictions and visualizations, not validation for vehicle deployment.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
13 October 2026
Original website title
taco-group/OpenEMMA · Official homepage title · retained in its source language
How we review sourcesSuggest a correction