PedAnalyze:

Pedestrian Behavior Annotation Tool

PedAnalyze is a Python-based video annotation tool to assist autonomous vehicle development and safety enhancement by capturing pedestrian and vehicle behaviors on the road, allowing researchers to use patterns to develop new technology.

Multi-Frame Annotations

Enables the analysis of pedestrian behavior across multiple video frames, using dashcam videos from YouTube/social media

Structured Datasets

Provides organized and labeled datasets for training and validating AV models. Easily accessible Python tool, JSON-ready for ML

Standardized Ontology

Captures human unpredictability through a consistent behavior model, and extensible to edge cases (atypical maneuvers)

Published in Institute of Electrical and Electronics Engineers (IEEE) Xplore: 15 July 2024

Presented at the 2024 IEEE Intelligent Vehicles Symposium (IV) on 02-05 June 2024 in Jeju Island, Republic of Korea

Used in research on accident analysis, AV simulation, and safety testing

IEEE Publication
ResearchGate Profile

Open-sourced the model architecture and inference workflow on GitHub

GitHub Repository
Next
Next

Research: ML Model