Hi,
I was wondering what the edge features for the VOC semantic image segmentation were? I see there are 3 dimensions, first is constant 1 (bias), channel 1 is between 0 and 1, and channel 2 seems to be between -pi/2 and pi/2, am I correct in assuming that this corresponds to some normalized distance and angle between superpixels?
Thank you, cheers
Hi,
I was wondering what the edge features for the VOC semantic image segmentation were? I see there are 3 dimensions, first is constant 1 (bias), channel 1 is between 0 and 1, and channel 2 seems to be between -pi/2 and pi/2, am I correct in assuming that this corresponds to some normalized distance and angle between superpixels?
Thank you, cheers