Hi! I’m the author of this blog and benchmark. You’re right. I’ll fix it in the ground-truth dataset. Thanks for pointing it out.