File size: 11,602 Bytes
9859ea2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 |
import numpy as np
import matplotlib.pyplot as plt
import torch
import torch.nn.functional as F
def get_next_click3D_torch_no_gt(prev_seg, img3D, threshold=170):
"""Selects prompt clicks from thresholded image (img3D) based on the previous segmentation (prev_seg).
Args:
prev_seg (torch.tensor): segmentation masks from previous iteration
img3D (torch.tensor): input images
threshold (int, optional): threshold value to apply to image for selecting point click. Defaults to 170.
Returns:
batch_points (list of torch.tensor): list of points to click
batch_labels (list of torch.tensor): list of labels corresponding to the points
NOTE: In this case, the labels are based on the thresholded image and not the ground truth.
"""
mask_threshold = 0.5
batch_points = []
batch_labels = []
pred_masks = prev_seg > mask_threshold
likely_masks = img3D > threshold # NOTE: Empirical threshold
fn_masks = torch.logical_and(likely_masks, torch.logical_not(pred_masks))
# NOTE: Given a strict/high threshold, the false positives are not going to be very useful (at least in my case)
# fp_masks = torch.logical_and(torch.logical_not(likely_masks), pred_masks)
for i in range(prev_seg.shape[0]): # , desc="generate points":
fn_points = torch.argwhere(fn_masks[i])
point = None
if len(fn_points) > 0:
point = fn_points[np.random.randint(len(fn_points))]
is_positive = True
# if no mask is given, random click a negative point
if point is None:
point = torch.Tensor(
[np.random.randint(sz) for sz in fn_masks[i].size()]
).to(torch.int64)
is_positive = False
bp = point[1:].clone().detach().reshape(1, 1, -1).to(pred_masks.device)
bl = (
torch.tensor(
[
int(is_positive),
]
)
.reshape(1, 1)
.to(pred_masks.device)
)
batch_points.append(bp)
batch_labels.append(bl)
return batch_points, batch_labels
def get_next_click3D_torch_no_gt_naive(prev_seg):
"""Selects prompt clicks from the area outside predicted masks based on previous segmentation (prev_seg).
Args:
prev_seg (torch.tensor): segmentation masks from previous iteration
Returns:
batch_points (list of torch.tensor): list of points to click
batch_labels (list of torch.tensor): list of labels corresponding to the points
NOTE: In this case, the labels are based on the predicted masks and not the ground truth.
"""
mask_threshold = 0.5
batch_points = []
batch_labels = []
pred_masks = prev_seg > mask_threshold
uncertain_masks = torch.logical_xor(
pred_masks, pred_masks
) # init with all False
for i in range(prev_seg.shape[0]):
uncertain_region = torch.logical_or(uncertain_masks[i, 0], pred_masks[i, 0])
points = torch.argwhere(uncertain_region) # select outside of pred mask
if len(points) > 0:
point = points[np.random.randint(len(points))]
is_positive = pred_masks[i, 0, point[1], point[2], point[3]]
bp = point[1:].clone().detach().reshape(1, 1, 3)
bl = torch.tensor([int(is_positive)], dtype=torch.long).reshape(1, 1)
batch_points.append(bp)
batch_labels.append(bl)
else:
point = torch.Tensor(
[np.random.randint(sz) for sz in pred_masks[i, 0].size()]
).to(torch.int64)
is_positive = pred_masks[i, 0, point[1], point[2], point[3]]
bp = point[1:].clone().detach().reshape(1, 1, 3)
bl = torch.tensor([int(is_positive)], dtype=torch.long).reshape(1, 1)
batch_points.append(bp)
batch_labels.append(bl)
return batch_points, batch_labels
def get_next_click3D_torch(prev_seg, gt_semantic_seg):
mask_threshold = 0.5
batch_points = []
batch_labels = []
# dice_list = []
pred_masks = prev_seg > mask_threshold
true_masks = gt_semantic_seg > 0
fn_masks = torch.logical_and(true_masks, torch.logical_not(pred_masks))
fp_masks = torch.logical_and(torch.logical_not(true_masks), pred_masks)
for i in range(gt_semantic_seg.shape[0]): # , desc="generate points":
fn_points = torch.argwhere(fn_masks[i])
fp_points = torch.argwhere(fp_masks[i])
point = None
if len(fn_points) > 0 and len(fp_points) > 0:
if np.random.random() > 0.5:
point = fn_points[np.random.randint(len(fn_points))]
is_positive = True
else:
point = fp_points[np.random.randint(len(fp_points))]
is_positive = False
elif len(fn_points) > 0:
point = fn_points[np.random.randint(len(fn_points))]
is_positive = True
elif len(fp_points) > 0:
point = fp_points[np.random.randint(len(fp_points))]
is_positive = False
# if no mask is given, random click a negative point
if point is None:
point = torch.Tensor(
[np.random.randint(sz) for sz in fn_masks[i].size()]
).to(torch.int64)
is_positive = False
bp = point[1:].clone().detach().reshape(1, 1, -1).to(pred_masks.device)
bl = (
torch.tensor(
[
int(is_positive),
]
)
.reshape(1, 1)
.to(pred_masks.device)
)
batch_points.append(bp)
batch_labels.append(bl)
return batch_points, batch_labels # , (sum(dice_list)/len(dice_list)).item()
import edt
def get_next_click3D_torch_ritm(prev_seg, gt_semantic_seg):
mask_threshold = 0.5
batch_points = []
batch_labels = []
# dice_list = []
pred_masks = prev_seg > mask_threshold
true_masks = gt_semantic_seg > 0
fn_masks = torch.logical_and(true_masks, torch.logical_not(pred_masks))
fp_masks = torch.logical_and(torch.logical_not(true_masks), pred_masks)
fn_mask_single = F.pad(fn_masks, (1, 1, 1, 1, 1, 1), "constant", value=0).to(
torch.uint8
)[0, 0]
fp_mask_single = F.pad(fp_masks, (1, 1, 1, 1, 1, 1), "constant", value=0).to(
torch.uint8
)[0, 0]
fn_mask_dt = torch.tensor(
edt.edt(fn_mask_single.cpu().numpy(), black_border=True, parallel=4)
)[1:-1, 1:-1, 1:-1]
fp_mask_dt = torch.tensor(
edt.edt(fp_mask_single.cpu().numpy(), black_border=True, parallel=4)
)[1:-1, 1:-1, 1:-1]
fn_max_dist = torch.max(fn_mask_dt)
fp_max_dist = torch.max(fp_mask_dt)
is_positive = (
fn_max_dist > fp_max_dist
) # the biggest area is selected to be interaction point
dt = fn_mask_dt if is_positive else fp_mask_dt
to_point_mask = dt > (max(fn_max_dist, fp_max_dist) / 2.0) # use a erosion area
to_point_mask = to_point_mask[None, None]
# import pdb; pdb.set_trace()
for i in range(gt_semantic_seg.shape[0]):
points = torch.argwhere(to_point_mask[i])
point = points[np.random.randint(len(points))]
if fn_masks[i, 0, point[1], point[2], point[3]]:
is_positive = True
else:
is_positive = False
bp = point[1:].clone().detach().reshape(1, 1, 3)
bl = torch.tensor(
[
int(is_positive),
]
).reshape(1, 1)
batch_points.append(bp)
batch_labels.append(bl)
return batch_points, batch_labels # , (sum(dice_list)/len(dice_list)).item()
def get_next_click3D_torch_2(prev_seg, gt_semantic_seg):
mask_threshold = 0.5
batch_points = []
batch_labels = []
# dice_list = []
pred_masks = prev_seg > mask_threshold
true_masks = gt_semantic_seg > 0
fn_masks = torch.logical_and(true_masks, torch.logical_not(pred_masks))
fp_masks = torch.logical_and(torch.logical_not(true_masks), pred_masks)
to_point_mask = torch.logical_or(fn_masks, fp_masks)
for i in range(gt_semantic_seg.shape[0]):
points = torch.argwhere(to_point_mask[i])
point = points[np.random.randint(len(points))]
# import pdb; pdb.set_trace()
if fn_masks[i, 0, point[1], point[2], point[3]]:
is_positive = True
else:
is_positive = False
bp = point[1:].clone().detach().reshape(1, 1, 3)
bl = torch.tensor(
[
int(is_positive),
]
).reshape(1, 1)
batch_points.append(bp)
batch_labels.append(bl)
return batch_points, batch_labels # , (sum(dice_list)/len(dice_list)).item()
def get_next_click3D_torch_with_dice(prev_seg, gt_semantic_seg):
def compute_dice(mask_pred, mask_gt):
mask_threshold = 0.5
mask_pred = mask_pred > mask_threshold
# mask_gt = mask_gt.astype(bool)
mask_gt = mask_gt > 0
volume_sum = mask_gt.sum() + mask_pred.sum()
if volume_sum == 0:
return np.NaN
volume_intersect = (mask_gt & mask_pred).sum()
return 2 * volume_intersect / volume_sum
mask_threshold = 0.5
batch_points = []
batch_labels = []
dice_list = []
pred_masks = prev_seg > mask_threshold
true_masks = gt_semantic_seg > 0
fn_masks = torch.logical_and(true_masks, torch.logical_not(pred_masks))
fp_masks = torch.logical_and(torch.logical_not(true_masks), pred_masks)
for i in range(gt_semantic_seg.shape[0]):
fn_points = torch.argwhere(fn_masks[i])
fp_points = torch.argwhere(fp_masks[i])
if len(fn_points) > 0 and len(fp_points) > 0:
if np.random.random() > 0.5:
point = fn_points[np.random.randint(len(fn_points))]
is_positive = True
else:
point = fp_points[np.random.randint(len(fp_points))]
is_positive = False
elif len(fn_points) > 0:
point = fn_points[np.random.randint(len(fn_points))]
is_positive = True
elif len(fp_points) > 0:
point = fp_points[np.random.randint(len(fp_points))]
is_positive = False
# bp = torch.tensor(point[1:]).reshape(1,1,3)
bp = point[1:].clone().detach().reshape(1, 1, 3)
bl = torch.tensor(
[
int(is_positive),
]
).reshape(1, 1)
batch_points.append(bp)
batch_labels.append(bl)
dice_list.append(compute_dice(pred_masks[i], true_masks[i]))
return batch_points, batch_labels, (sum(dice_list) / len(dice_list)).item()
def show_mask(mask, ax, random_color=False):
if random_color:
color = np.concatenate([np.random.random(3), np.array([0.6])], axis=0)
else:
color = np.array([251 / 255, 252 / 255, 30 / 255, 0.6])
h, w = mask.shape[-2:]
mask_image = mask.reshape(h, w, 1) * color.reshape(1, 1, -1)
ax.imshow(mask_image)
def show_point(point, label, ax):
if label == 0:
ax.add_patch(plt.Circle((point[1], point[0]), 1, color="red"))
else:
ax.add_patch(plt.Circle((point[1], point[0]), 1, color="green"))
# plt.scatter(point[0], point[1], label=label)
if __name__ == "__main__":
gt2D = torch.randn((2, 1, 256, 256)).cuda()
prev_masks = torch.zeros_like(gt2D).to(gt2D.device)
batch_points, batch_labels = get_next_click3D_torch(
prev_masks.to(gt2D.device), gt2D
)
print(batch_points)
|