Thanks for your contributions and for releasing the dataset.
I am conducting a follow-up research on the M3D paper and after reading the content, I could not find details about the training samples used in training, given the huge amount of images (~100K samples x ~300/sample = 3M images).
Since the test samples used are quite small, did you use all the samples or just part of them for training? If only a portion of the training samples were used, could you also release the corresponding IDs so that I can keep these samples?
Hi,
Thanks for your contributions and for releasing the dataset.
I am conducting a follow-up research on the M3D paper and after reading the content, I could not find details about the training samples used in training, given the huge amount of images (~100K samples x ~300/sample = 3M images).
Since the test samples used are quite small, did you use all the samples or just part of them for training? If only a portion of the training samples were used, could you also release the corresponding IDs so that I can keep these samples?
Best, Ethan