Question about the training process
#2
by omarom - opened
Hi, thank you for sharing your work!
I'm curious about your training process, could you please elaborate on it? Since, given an input image, there are multiple possible objects to remove, it feels like the training objective is not well defined. How does your training dataset look?
Hey @omarom ,
Thanks for asking about this in detail. Your questions are welcome.
Yes, the dataset may contain a mixed combination of single and multiple object removal target images, and the majority of the training dataset consists of real-world based images.
*the preview content of the dataset cannot be disclosed due to certain reasons.
for example, consider the following case, as it contains a clear example of multiple items being removed.
prithivMLmods changed discussion status to closed