Training images of different dimensions (and additional questions on config & custom training)

[riven314]

Thanks for the great work! I have 2 questions:

1. I wanna do custom training on images of different dimensions, could your training pipeline handle that? or I needa write my own logic to address them?
2. I saw min_dim in the config. What does it actually mean?
3. for custom training, in general what are the configurations I may need to tweak?

Thank you so much!