| import drjit as dr |
| import mitsuba as mi |
| import numpy as np |
| from PDE3D import ArrayXf |
|
|
| def MSE(img , img_ref = 0): |
| val = img |
| val_ref = img_ref |
| if isinstance(val, mi.TensorXf): |
| val = val.array |
| if isinstance(val_ref, mi.TensorXf): |
| val_ref = val_ref.array |
| return dr.block_sum(dr.square(val - val_ref), dr.width(val)) / dr.width(val) |
|
|
|
|
| def MSE_numpy(val :np.array , val_ref : np.array = 0): |
| return np.sum(np.square(val - val_ref), axis = tuple(range(1, val.ndim))) / (np.size(val) / val.shape[0]) |
|
|
| def MSE_vol(img , img_ref = 0): |
| return dr.sum(dr.square(img - img_ref).array) / (img.shape[1] * img.shape[2] * img.shape[3]) |
|
|
| def MSE_slice(img , img_ref = 0): |
| return dr.sum(dr.square(img - img_ref).array) / (img.shape[1] * img.shape[2]) |
|
|
| def compute_loss_grad_slice(result, result_ref = 0): |
| return (2 * (result - result_ref)) / (result.shape[1] * result.shape[2]) |
|
|
| def compute_dL_slice(loss_grad, spp): |
| size = loss_grad.shape[1] * loss_grad.shape[2] * spp |
| dL = dr.zeros(ArrayXf, shape = (loss_grad.shape[0], size)) |
| for i in range(loss_grad.shape[0]): |
| dL[i] = dr.repeat(loss_grad[i].array, spp) / spp |
| return dL |
|
|
| def compute_loss_grad_vol(result, result_ref = 0): |
| return (2 * (result - result_ref)) / (result.shape[1] * result.shape[2] * result.shape[3]) |
|
|
| def compute_dL_vol(loss_grad, spp): |
| size = loss_grad.shape[1] * loss_grad.shape[2] * loss_grad.shape[3] * spp |
| dL = dr.zeros(ArrayXf, shape = (loss_grad.shape[0], size)) |
| for i in range(loss_grad.shape[0]): |
| dL[i] = dr.repeat(loss_grad[i].array, spp) / spp |
| return dL |
