0

In detectron2 there seems to be two settings related to batch size in the config file.

They are

SOLVER.IMS_PER_BATCH and MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE

I have seen both being used as batch size online and not sure how they differ. What is the difference and if they work together, how?

devman3211
  • 63
  • 1
  • 11

1 Answers1

1

SOLVER.IMS_PER_BATCH is what's commonly known as batch size in deep learning and refers to the number of training examples utilized in one iteration.

MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE is a parameter that is used to sample a subset of proposals coming out of RPN to calculate cls and reg loss during training.

Kami
  • 164
  • 11