Ask Your Question

Best way to extract synchronized data from bag file for deep learning traning

asked 2018-08-06 14:57:22 -0500

AutoCar gravatar image

Hi, I have a MIT-RACECAR and would like to use deep learning to drive the car autonomously. The neural network takes image and odom as input and gives ackerman message as output. To train the network, I have driven the car with teleop and collected a bag file. Now I need to prepare the data for my network. What is the best way to extracted paired image, odom, and ackerman message from the bag file so I can use it for deep learning training?

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted

answered 2018-08-06 16:09:04 -0500

The question is how similar are the frequencies and offsets of the three message topics? If they're high frequency then you may be able to use the message_filters package to do what you want very easily. You'll probably need the policy based synchronizer

However if say your akerman steering messages are at a much lower frequency than your images and odometry for example, then you may want to interpolate the ackerman messages to the exact time of the image. Or you may need to interpolate both the ackerman and the odometry topics on this way to match the images. This interpolation method is more complicated but will give more precise results.

Which of these you choose will depend on the rate of each of the three topics and how precisely you need them to be aligned in time.

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools

1 follower


Asked: 2018-08-06 14:57:22 -0500

Seen: 288 times

Last updated: Aug 06 '18