One Step Right, One Step Down

A network flattens $6 \times 6$ grayscale images row by row into $36$ numbers. Rows and columns are numbered $0$ to $5$.

Image A is black everywhere except a vertical bar of brightness $200$ filling column $2$ from top to bottom:

10  0  200  0  0  0
20  0  200  0  0  0
30  0  200  0  0  0
40  0  200  0  0  0
50  0  200  0  0  0
60  0  200  0  0  0

One unit in the first ordinary layer has ended up with weight $1$ on the six inputs that came from column $2$, and weight $0$ on the other $30$. It outputs the plain weighted sum of its $36$ inputs, with no bias and no activation, so on Image A it outputs $6 \times 200 = 1{,}200$.

The same unit is then shown three more images:

• Image B: Image A slid one column to the right.
• Image C: Image A slid one row down.
• Image D: black everywhere except a horizontal bar of brightness $200$ filling row $3$ from left to right.

When an image slides, whatever is pushed past the border is lost, and the strip that opens up on the other side is filled with $0$.

What does the unit output for B, C and D?