I have found in model/base.py
the direction prediction are using up1_embedding:
if self.direction_pred: x_direction = self.up1_embedded(x2, x1) x_direction = self.up2_direction(x_direction)
|
x_direction = self.up1_embedded(x2, x1) |
why using up_1embedded blocks instead of using up1_direction as the up1_direction is already defined.
and will the weights of up1_embedded blocks are affected by the direction prediction?
I have found in model/base.py
the direction prediction are using up1_embedding:
if self.direction_pred: x_direction = self.up1_embedded(x2, x1) x_direction = self.up2_direction(x_direction)HDMapNet/model/base.py
Line 131 in 7911227
why using up_1embedded blocks instead of using up1_direction as the up1_direction is already defined.
and will the weights of up1_embedded blocks are affected by the direction prediction?