JierunChen
A large pre-training weight
Hi, thanks for your interest in the work. If you are referring 128, 256, 512, and 1048 as the network widths of four stages in a single network, it has been introduced and trained as FasterNet-S indeed. You may also check out the larger FasterNet-M and FasterNet-L. For training even huger models, it's currently not on our to-do list.
I hope you can launch such a deeper network of 128,256,512,1048 dimensions, so that I can compare with other similar large networks to find out how effective your network is, if your network of this dimension is relatively shallow, it may not be as effective as other similar networks, and in the detection of the neck part is required 128,256,512,1048 this dimension, unfortunately I don't have a pre-trained machine that can implement this large model, otherwise I can implement it myself