Should the azimuth attribute values not be between 0 and 180 degrees? Normally azimuth only gives the orientation, not the direction.

The azimuth angle is the direction of maximum dip, relative to the reference direction (usually geographical North or inline/sail direction). In other words, it is the direction that a ball would roll if it was located on the point in space. The normal vector of a tangent plane can point to all possible directions, and the valid orientation range will hence be 0-360 degrees. If we would only allow azimuth angles to be in a half sphere (0-180 degrees) then the dip angle would need to be signed in order to allow all possible degrees of freedom. But, by convention, dip angles are unsigned. This is consistent with general practice in the industry and is, in particular, consistent with the convention defined by Prof. Kurt Marfurt, a recognized expert in the field (see «Kurt Marfurt's paper» for detail).

Why is the middle region of azimuth ground truth in the dataset snapshot page zero? The seismic data loaded seems to have valid values in this region. And the orientation is also changing within this region, so should it not be zero?

This is because the middle region is a fault block, with a different azimuth direction than the seismic reflectors to the left and right of the two faults delineating the fault block in the center. The azimuth angle of the center fault block is 0 degrees relative to the inline direction. Outside the fault block the azimuth angle is 180 degrees for this particular inline.

Figure: Inline 76 (this is the center inline) going through the seismic cube (left) and through the dip azimuth inline “ground truth” cube (right).