MonsterBorg driving around obstacles

As many of you have found out, avoiding other robots in Formula Pi is hard. In order to do well in the challenge rounds improving the avoidance code is an absolute must!

As it is not possible to get lots of images from the robots in different situations during a test session we have decided to help everyone out by recording a dataset with lots of traffic to avoid. This can be used to both understand what the MonsterBorg sees while racing around stationary robots.

The stationary robots are using the five colours we use during the racing so that you can see how this affects the code as well and tune things as you see fit. You might even be able to use the data to train a neural network if you want.

Here is the raw footage we have taken from the MonsterBorg driving around other robots under human control:

Note that it is rotated 180 degrees because the Pi camera is upside-down in a MonsterBorg.

You can download the full dataset from these links:

• Raw recorded footage - H264
• Playable video file - MP4
• Individual frame by frame images - JPEG

The images should work with the SimulationImages.py script for testing your own code. Alternatively the video files can be read from directly by OpenCV on many platforms.

We also have an earlier dataset which has no robots on the track here. This includes other conditions such as flipping over and hitting the wall. It may also be useful to test your code for false positives.

Stay tuned for our next post where we explain how we captured these images :)

Update:
See how we recorded the images in the following blog post: Recording images from MonsterBorg .