EuRoC Forum

Dear organizers,

are the following two assumptions correct about tasks 1 and 2?

• The final evaluation will be performed using exactly the same bagfiles you published in conjunction with the virtual machines.
• Contestants are not required to submit or publish their full source code for solving the tasks.

In case both are true I wonder: How can you effectively prevent teams from cheating? I could imagine teams precomputing very accurate solutions to both tasks using much more processing power and time than available on the final evaluation hardware and afterwards hardcoding this accurate result into their solution program. Result: Perfect computational efficiency, perfect accuracy.

What is your stance on that?

Thanks in advance,
Sebastian

Dear Sebastian,

we're currently preparing the instructions for the final evaluation, which will also clarify on this.
1) the final dataset will not be the one in the current VM. It is however recorded in the same area, we just did a second flight of approx. the same path. So the path changes, but the overall conditions (lighting, texture) are approximately the same. We will most likely publish a sequence of the beginning of each dataset, such that contestants can initialize their solution and make sure everything is going to run properly.

2) contestants submit the VM with their solutions, where they can remove their source code. Just make sure that your solution is still executable with the provided launch files.

Best, Markus

Markus, thank you very much for clearing that up.

Hi,

I was wondering whether you still intend to do this:

markusachtelik wrote:We will most likely publish a sequence of the beginning of each dataset, such that contestants can initialize their solution and make sure everything is going to run properly.

Thanks in advance,
Sebastian