Trimming Data Tubs

Proper data preparation is essential for training a reliable model. Data tubs contain the collected driving data, but not all data points are useful for training. Here’s how to trim data tubs to enhance model training:

• Reviewing Data: Use the Donkey Car management tool to review the images and corresponding driving data collected during your sessions.
• Identifying Useful Data: Look for data that represents good driving behavior. Remove any images or sequences where the car is off-track, stationary without purpose, or involved in a collision.
• Trimming Command: Utilize the command donkey tubclean --tub <path_to_tub> to launch a web interface for easier data selection and trimming.