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

Trajectron++

Trajectron++ is an MIT-licensed research-code repository for forecasting dynamically feasible trajectories from heterogeneous data, accompanying an ECCV 2020 paper. It includes documented data-processing and model-training workflows for ETH and UCY pedestrian datasets and nuScenes.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Trajectron++ — 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 the repository when you need research code and documented data-preparation and training workflows for heterogeneous trajectory forecasting, rather than a hosted prediction application.

Who it is for

  • Researchers reproducing or extending trajectory-forecasting experiments.
  • Teams studying pedestrian movement or autonomous-driving prediction scenarios.
  • Practitioners who need to prepare supported datasets and train models using the repository's documented workflow.

Documented uses

  • Process included ETH and UCY pedestrian splits into the model's data format.
  • Prepare nuScenes data and its map expansion pack for model use.
  • Train pedestrian-dataset models with selected training and evaluation data, configurations, and logging options.
  • Compare research runs using configured evaluation intervals and TensorBoard visualizations.
Before you choose

Requirements & limits

The repository is public and specifies an MIT license. Processing nuScenes data requires signing up with nuScenes.

  • The documented environment uses Conda, Python 3.6, pip-installed requirements, and an IPython kernel.
  • The README recommends cloning submodules and notes that the master branch can differ from the ECCV 2020 version, including possible interoperability differences.
  • ETH and UCY preprocessed splits are included; nuScenes data and its map expansion pack must be obtained separately.
  • The README points to a newer codebase in NVlabs/adaptive-prediction.

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 ETH and UCY pedestrian data

Convert the repository's included preprocessed pedestrian splits into the model's data format.

Before you start

  • Clone the repository with its submodules, or initialize and update the submodules after cloning.
  • Create the documented Conda environment with Python 3.6 and install dependencies from requirements.txt.
  • The repository's preprocessed ETH and UCY splits are in experiments/pedestrians/raw.

Steps

  1. Set up the Conda environment and install the listed requirements.

  2. Change to the experiments/pedestrians directory.

  3. Run python process_data.py.

Expected outcome

Pedestrian data is processed into a format the model can use.

Guide limitations

  • The README estimates processing at around 10–15 minutes, depending on the computer.
  • This guide covers data preparation, not model training or execution.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
6 November 2026
Original website title
StanfordASL/Trajectron-plus-plus · Official homepage title · retained in its source language
How we review sourcesSuggest a correction