Traffic Light Detection: Generate labeled images for classifier.

Important: Please also include the names and Udacity account emails of you and your team members in the Notes to reviewer when uploading your submission. Without these, we will not be able to run the code on Carla.

This node listens to /current_pose and /base_waypoints to understand where the vehicle is in the world. Then it determines what traffic lights are nearby, based on their permanent location in the map (stored in traffic_light_config).

The node determines the color of the upcoming light by listening to /camera/image_raw.

Finally, the node publishes the location of any upcoming red lights to /traffic_waypoint, so that the vehicle's path can be adjusted by waypoint_updater.

Traffic Light Detection: Analyse how labeled image can be generated/Recorded from /camera/image_raw and /vehicle/traffic_lights.

We can also consider using MPC

Important: Please leave the Notes to reviewer blank when uploading your submission.

For know it's only brute force.

Add Udacity account emails to README

Subscribe to /twist_cmd and use various controllers to provide appropriate throttle, brake, and steering commands.

Can be realized with the /vehicle/traffic_lights topic. Does not use images. Returns Next Waypoint with red traffic light.

This will help us determine what data to collect for training the TL model

Implement the closest waypoint algorithm for a given pose in tl_detector.py get_closest_waypoint()

Traffic Light Detection: Make camera images from /camera/image_raw topic visible with the rqt_image_view command.
Currently there is an error message: ImageView.callback_image() could not convert image from '8UC3' to 'rgb8' ([8UC3] is not a color format. but [rgb8] is. The conversion does not make sense)

Implement the closest waypoint algorithm for a given pose in tl_detector.py get_closest_waypoint()

It's mentioned in the architecture and in the video summary, but it doesn't have any requirement or code snippet!

Subscribe to the topics

• /base_waypoints
• /current_pose

and publish a list of waypoints to

• /final_waypoints

Maybe it would be useful to create a seperate project/repository for this.

Traffic Light Detection: Determine which classifier architecture (e.g. YOLO) is appropriate to the Traffic Light Problem.

Once traffic light detection is working properly, you can incorporate the traffic light data into your waypoint updater node. To do this, you will need to add a subscriber for the /traffic_waypoint topic and implement the traffic_cb callback for this subscriber.