Need for Speed SDC Capstone Project

This is the project repo for the final project of the Udacity Self-Driving Car Nanodegree: Programming a Real Self-Driving Car. For more information about the project, see the project introduction here.

Need for Speed is comprised of the following engineers:

• Kiarie Ndegwa ([email protected])
• Joseph Zhou ([email protected])
• Pramod BM ([email protected])
• Sidharth Varier ([email protected])
• Mike Rzucidlo ([email protected])

To build the environment needed to run the code in this repo, you should follow the instructions in the original Udacity project instructions.

Building and running the project

The models for traffic light classification are shared in the google drive. The link is given in the IMPORTANT NOTE section The models used in this exercise are based on the Faster rcnn resnet 101 architecture; pretrained on the Coco dataset and fine-tuned on the Bosch traffic signal dataset.

To use the models in the simulator or carla you must:

• Download and extract the models from the link. (model.tar.gz contains two models; one trained for the simulator and the other for the real world.)
• The respective files are labelledfrozen_inference_graph_real.pb and frozen_inference_graph_sim.pb
• Copy both frozen models into ros/src/tl_detector/light_classification/model into your local repo

NOTE: The zip submission already contains the models. So even if the above steps are not carried out, the project should still work.

Once set up you need to go to the folder labeled NFS-Capstone/ros and type in the following and build commands:

• catkin_make
• devel/setup.sh
• roslaunch launch/styx.launch