This repository contains code and annotations for the Pedestrian Intention Estimation (PIE) dataset: A. Rasouli, I. Kotseruba, T. Kunic, J.K. Tsotsos, PIE: A Large-Scale Dataset and Models for Pedestrian Intention Estimation, ICCV, 2019 and a series of scripts for visualization and scenario evaluation.
Videos are grouped into 6 sets corresponding to different routes driven in Toronto, Canada.
The total size of all video clips is approx. 74GB.
Download links YorkU server Google Drive
- annotations: Contains the annotations and a script for extracting images from raw videos
- scenarioEval: Contains code for scenario generation and metrics for trajectory and action prediction
- visualization: Contains a series of scripts for visualizing data samples and illustrating trajectory and action prediction modules
- model_outputs: Contains models' outputs for behavior prediction to be used for evaluation and visualization
- utilities: Contains configuration file for evaluation and visualization, dataset interfaces, and other utility functions
- camera_params: Contains camera parameters for the PIE dataset
If you use our dataset, please cite:
@InProceedings{Rasouli_2019_ICCV,
author = {Rasouli, Amir and Kotseruba, Iuliia and Kunic, Toni and Tsotsos, John K.},
title = {PIE: A Large-Scale Dataset and Models for Pedestrian Intention Estimation and Trajectory Prediction},
booktitle = {International Conference on Computer Vision (ICCV)},
year = {2019}}
Please send an email to [email protected] or [email protected] if there are any problems with downloading or using the data.
This project is licensed under the MIT License - see the LICENSE file for details