A projection shadow sums all pixel values along one line of sight, turning a 2-D grid into a 1-D measurement. The row with cells 1, 2, 3 casts a shadow of 1 + 2 + 3 = 6 while the column sums are 1, 2, and 3 exactly for the stated toy projection ledger; real imaging measurements have noise, calibration limits, sampling geometry, attenuation and scattering, and measurement uncertainty.
A one-row training image turns pixel indices into one checked sum. Exact arithmetic here means exact results for the stated model inputs; measured inputs still carry uncertainty and significant-figure limits.
highlighted = computed this step
Read a one-row image
The row contains pixel indices 1, 2, and 3.
1+2+3
A row shadow is a line sum
The projection score is 6, because the whole row lies under the same ray.