PedAnalyze-Web is a web-based interface for the PedAnalyze tool, designed to streamline the annotation and analysis of pedestrian behavior in dash-cam videos. It is built to support autonomous vehicle research. It enables researchers, developers, and data scientists to create structured datasets with pre-defined tags for pedestrian and vehicle interactions, capturing critical scenarios like accidents or near-misses. With support for single-frame and multi-frame annotations, PedAnalyze-Web reduces redundancy and enhances temporal relationship analysis, making it a valuable tool for advancing safer autonomous driving systems.
This project complements the PedAnalyze system, enabling researchers, traffic analysts, and AV developers to create high-quality behavior datasets for intelligent systems.
For full usage, setup, and annotation guidelines, visit:
👉 https://pedanalyze.readthedocs.io/en/latest/
Check this project out at https://pedanalyze-web.vercel.app/
- 🎥 Video Integration: Load videos from YouTube URLs.
- ⏱ Precise Frame Control: Use a timestamp input for frame-level accuracy.
- 🏷 Behavioral Tagging: Add structured tags for pedestrians, vehicles, and environmental context.
- 💾 JSON Export: Save annotations in a clean JSON format, ready for modeling or evaluation.
- 🌐 Web-First UI: Built with React for smooth interaction and low-latency annotation sessions.
- 🚦 Annotating pedestrian interactions in traffic surveillance footage
- 🤖 Creating training data for behavior prediction in autonomous vehicle systems
- 🧍 Studying human movement, attention, and trajectory in urban settings
- 🧪 Generating fine-grained datasets for ML model evaluation and simulation environments
git clone https://github.com/PedSim/pedanalyze-web.git
cd pedanalyze-web
docker build -t pedanalyze-web .
docker run -p 3000:3000 pedanalyze-webgit clone https://github.com/PedSim/pedanalyze-web.git
cd pedanalyze-web
npm install
npm startIf you use this tool for academic purposes, please cite the following
@inproceedings{inproceedings,
author = {Huang, Taorui and Muktadir, Golam Md and Sripada, Srishti and Saravanan, Rishi and Yuan, Amelia and Whitehead, Jim},
year = {2024},
month = {03},
pages = {},
title = {PedAnalyze - Pedestrian Behavior Annotator and Ontology},
doi = {10.1109/IV55156.2024.10588755}
}This is our ongoing work identifying the archetypes using this tool:
@inproceedings{inproceedings,
author = {Muktadir, Golam Md and Huang, Taorui and Bansal, Ritvik and Gaidhani, Namita and Jubaer, S M and Lin, Michael and Whitehead, Jim},
year = {2025},
month = {03},
pages = {},
title = {Pedestrian Archetypes - The Must-Have Pedestrian Models for Autonomous Vehicle Safety Testing}
}We’d love to hear from you!
If you have any questions, feature requests, or want to contribute, feel free to reach out via:
- 📧 Email: [email protected]
