Pedestrian Detection & Tracking

Universitat Politècnica de Catalunya (UPC), Barcelona Faculty of Informatics (FIB), Cognitive Interaction with Robots (MAI-CIR), Final Course Project WS 2018/2019

Description

This is an integration of YOLO object detection with SORT tracking algorithm modified for pedestrian detection and tracking applied to the JAAD dataset.

For more detail on the project and references, please have a look into the final report

How to

The project is split into several subtasks with a self-contained Google Colab Jupyter Notebook.

1. Run the JAAD Dataset Preprocessing Notebook
2. Run the YOLO Pedestrian Detection Notebook to train the neural network
   • you will probably exceed the Google Colab runtime limit so you can use the Save Checkpoints Notebook to create checkpoints to restart training from
3. Run the SORT Tracking on your detection results