Hi,
I am trying to perform incremental training on VITMAE using my own dataset which have around 150 images.
I am just perfroming VITAutoEncoder model training and have explore all the three variants with different number of epochs.
I have use transfer learning, i.e. used pretrained weights and does incremental training and trained from the random weights state also.
Issue that i am facing is generated images grids are not smooth and can clearly see boundaries on each sub block of the image.
I am not sure how to fix this issue. I am exploring overlapping grids, looking forward for expert support on this.