Open ChenhanXmu opened 7 years ago
In detector step, why the input dimension are different in training (1L,128L, 128L, 128L) )and testing (1L, 208L, 208L, 208L))? It's mean the input of the network is variable?
Yes, it is variable, but recently we found that the changing of shape may lead to unwanted changing of output. Please be careful about this
In detector step, why the input dimension are different in training (1L,128L, 128L, 128L) )and testing (1L, 208L, 208L, 208L))? It's mean the input of the network is variable?