Spaces:
Running
Running
| import numpy as np | |
| def pad_reflect(image, pad_size): | |
| imsize = image.shape | |
| height, width = imsize[:2] | |
| new_img = np.zeros([height+pad_size*2, width+pad_size*2, imsize[2]]).astype(np.uint8) | |
| new_img[pad_size:-pad_size, pad_size:-pad_size, :] = image | |
| new_img[0:pad_size, pad_size:-pad_size, :] = np.flip(image[0:pad_size, :, :], axis=0) | |
| new_img[-pad_size:, pad_size:-pad_size, :] = np.flip(image[-pad_size:, :, :], axis=0) | |
| new_img[:, 0:pad_size, :] = np.flip(new_img[:, pad_size:pad_size*2, :], axis=1) | |
| new_img[:, -pad_size:, :] = np.flip(new_img[:, -pad_size*2:-pad_size, :], axis=1) | |
| return new_img | |
| def unpad_image(image, pad_size): | |
| return image[pad_size:-pad_size, pad_size:-pad_size, :] | |
| def unpad_patches(image_patches, padding_size): | |
| return image_patches[:, padding_size:-padding_size, padding_size:-padding_size, :] | |
| def split_image_into_overlapping_patches(image_array, patch_size, padding_size=2): | |
| xmax, ymax, _ = image_array.shape | |
| x_extend = (patch_size - (xmax % patch_size)) % patch_size | |
| y_extend = (patch_size - (ymax % patch_size)) % patch_size | |
| extended_image = np.pad(image_array, ((0, x_extend), (0, y_extend), (0, 0)), 'edge') | |
| padded_image = np.pad(extended_image, ((padding_size, padding_size), (padding_size, padding_size), (0, 0)), 'edge') | |
| xmax_pad, ymax_pad, _ = padded_image.shape | |
| patches = [] | |
| # PERBAIKAN 1: Hitung jumlah patch per baris secara akurat dari loop | |
| n_patches_per_row = len(range(padding_size, ymax_pad - padding_size, patch_size)) | |
| for x in range(padding_size, xmax_pad - padding_size, patch_size): | |
| for y in range(padding_size, ymax_pad - padding_size, patch_size): | |
| patch = padded_image[x-padding_size:x+patch_size+padding_size, y-padding_size:y+patch_size+padding_size, :] | |
| patches.append(patch) | |
| return np.array(patches), padded_image.shape, n_patches_per_row | |
| def stich_together(patches, padded_image_shape, target_shape, padding_size=4, n_patches_per_row=None): | |
| xmax, ymax, _ = padded_image_shape | |
| patches = unpad_patches(patches, padding_size) | |
| patch_size = patches.shape[1] | |
| if n_patches_per_row is None: | |
| n_patches_per_row = max(1, round(ymax / patch_size)) | |
| complete_image = np.zeros((xmax, ymax, 3)) | |
| row = -1 | |
| col = 0 | |
| for i in range(len(patches)): | |
| if i % n_patches_per_row == 0: | |
| row += 1 | |
| col = 0 | |
| # PERBAIKAN 2: Gunakan min() untuk mencegah assignment melebihi batas array (mencegah broadcast error) | |
| x_start = row * patch_size | |
| x_end = min(x_start + patch_size, xmax) | |
| y_start = col * patch_size | |
| y_end = min(y_start + patch_size, ymax) | |
| patch_x_end = x_end - x_start | |
| patch_y_end = y_end - y_start | |
| complete_image[x_start:x_end, y_start:y_end, :] = patches[i][:patch_x_end, :patch_y_end, :] | |
| col += 1 | |
| return complete_image[0: target_shape[0], 0: target_shape[1], :] |