pydriver 1.0.1

PyDriver is a Python (2.7, 3.x) framework for training and evaluating object detectors and classifiers in road traffic environment. It aims at researchers experimenting with learning based 3D object detection, classification and tracking.

PyDriver helps you with following tasks:

• read data from standard datasets (currently integrated: KITTI)

• reconstruct 3D scene from stereo images, detect/remove ground plane, restrict viewport etc.

• detect 3D keypoints and extract 3D feature descriptors

• train classifiers available in scikit-learn (support vector machines, decision trees and many more)

• visualize point clouds, extracted keypoints and detections using Point Cloud Library

• evaluate produced detections with respect to dataset ground truth

Of course, you can write your own modules and enhancements such as interfaces to your custom dataset or implement novel feature descriptors. The point of PyDriver is to provide a basic pipeline for object detection which you can enhance with your own algorithms without reimplementing the routine stuff.

PyDriver is subject to the MIT licence.

PyDriver installation documentation

PyDriver documentation

The most current development version is available here: http://github.com/lpltk/pydriver