Marcus Loo Vergara
All projects

Machine Learning

Autonomous Vehicle Perception

Faster R-CNN trained on the Udacity driving dataset to detect cars, trucks, pedestrians, bikers and traffic lights.

Source on GitHub
Street scene with detected pedestrians, cars and a traffic light outlined
Road with several detected cars and a traffic light outlined
Tree-lined road with a detected truck and car outlined

1 / 3

For my project in TDT4265: Computer Vision at NTNU, I trained Faster R-CNN on the Udacity dataset. It detects five classes with moderate accuracy: cars, trucks, pedestrians, traffic lights and bikers.

It is written in Python 3 using pycaffe, the Python interface to the Caffe deep learning framework.