Pedestrian Tracker demo

This project demonstrates a simple pedestrian tracker using HOG features support vector machine and a particle filter. It is not very fast as it runs a HOG detection for every particle, every frame.

Files

• trainer.py: run this to train a SVM on your training dataset
• tracker.py: track pedestrians in your test data using the previously trained SVM
• ParticleFilter.py: class that does the actual tracking
• dataset.py/dalal.py/iccv07.py: scripts that read in different datasets

library dependencies

• OpenCV2
• numpy
• matplotlib
• sklearn

datasets

• ICCV07 Pedestrian dataset: http://www.vision.ee.ethz.ch/datasets_extra/iccv07-data.tar.gz
• DALAL Pedestrian dataset

Screenshots