Hi!
Currently, the depth control model accepts uint8 input with specific encoding rules which convert the real distance value into a grayscale value. I wonder:
1). Will the real-distance input improve the structural control(more accurate...etc) of generation?
2). Is there any suggestion for doing so?
Hi! Currently, the depth control model accepts uint8 input with specific encoding rules which convert the real distance value into a grayscale value. I wonder: 1). Will the real-distance input improve the structural control(more accurate...etc) of generation? 2). Is there any suggestion for doing so?