I think the structure of fusion net in prototxt is not exactly the same with the structure announced in your paper. I mean the 'fuse_interconv1' layer, aka 'rconv1' in Table 4.
prototxt:
paper_supp:
BTW, is '6/64' of conv0's Ch I/O a mistake in Table 4?
I think the structure of fusion net in prototxt is not exactly the same with the structure announced in your paper. I mean the 'fuse_interconv1' layer, aka 'rconv1' in Table 4. prototxt:
paper_supp:
BTW, is '6/64' of conv0's Ch I/O a mistake in Table 4?