Spaces:
Runtime error
Runtime error
File size: 21,344 Bytes
b334e29 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 |
import torch
import torch.nn as nn
import torch.nn.functional as F
from mmcv.runner import auto_fp16, force_fp32
from torch.nn.modules.utils import _pair
from mmdet.core import build_bbox_coder, multi_apply, multiclass_nms
from mmdet.models.builder import HEADS, build_loss
from mmdet.models.losses import accuracy
@HEADS.register_module()
class BBoxHead(nn.Module):
"""Simplest RoI head, with only two fc layers for classification and
regression respectively."""
def __init__(self,
with_avg_pool=False,
with_cls=True,
with_reg=True,
roi_feat_size=7,
in_channels=256,
num_classes=80,
bbox_coder=dict(
type='DeltaXYWHBBoxCoder',
clip_border=True,
target_means=[0., 0., 0., 0.],
target_stds=[0.1, 0.1, 0.2, 0.2]),
reg_class_agnostic=False,
reg_decoded_bbox=False,
loss_cls=dict(
type='CrossEntropyLoss',
use_sigmoid=False,
loss_weight=1.0),
loss_bbox=dict(
type='SmoothL1Loss', beta=1.0, loss_weight=1.0)):
super(BBoxHead, self).__init__()
assert with_cls or with_reg
self.with_avg_pool = with_avg_pool
self.with_cls = with_cls
self.with_reg = with_reg
self.roi_feat_size = _pair(roi_feat_size)
self.roi_feat_area = self.roi_feat_size[0] * self.roi_feat_size[1]
self.in_channels = in_channels
self.num_classes = num_classes
self.reg_class_agnostic = reg_class_agnostic
self.reg_decoded_bbox = reg_decoded_bbox
self.fp16_enabled = False
self.bbox_coder = build_bbox_coder(bbox_coder)
self.loss_cls = build_loss(loss_cls)
self.loss_bbox = build_loss(loss_bbox)
in_channels = self.in_channels
if self.with_avg_pool:
self.avg_pool = nn.AvgPool2d(self.roi_feat_size)
else:
in_channels *= self.roi_feat_area
if self.with_cls:
# need to add background class
self.fc_cls = nn.Linear(in_channels, num_classes + 1)
if self.with_reg:
out_dim_reg = 4 if reg_class_agnostic else 4 * num_classes
self.fc_reg = nn.Linear(in_channels, out_dim_reg)
self.debug_imgs = None
def init_weights(self):
# conv layers are already initialized by ConvModule
if self.with_cls:
nn.init.normal_(self.fc_cls.weight, 0, 0.01)
nn.init.constant_(self.fc_cls.bias, 0)
if self.with_reg:
nn.init.normal_(self.fc_reg.weight, 0, 0.001)
nn.init.constant_(self.fc_reg.bias, 0)
@auto_fp16()
def forward(self, x):
if self.with_avg_pool:
x = self.avg_pool(x)
x = x.view(x.size(0), -1)
cls_score = self.fc_cls(x) if self.with_cls else None
bbox_pred = self.fc_reg(x) if self.with_reg else None
return cls_score, bbox_pred
def _get_target_single(self, pos_bboxes, neg_bboxes, pos_gt_bboxes,
pos_gt_labels, cfg):
"""Calculate the ground truth for proposals in the single image
according to the sampling results.
Args:
pos_bboxes (Tensor): Contains all the positive boxes,
has shape (num_pos, 4), the last dimension 4
represents [tl_x, tl_y, br_x, br_y].
neg_bboxes (Tensor): Contains all the negative boxes,
has shape (num_neg, 4), the last dimension 4
represents [tl_x, tl_y, br_x, br_y].
pos_gt_bboxes (Tensor): Contains all the gt_boxes,
has shape (num_gt, 4), the last dimension 4
represents [tl_x, tl_y, br_x, br_y].
pos_gt_labels (Tensor): Contains all the gt_labels,
has shape (num_gt).
cfg (obj:`ConfigDict`): `train_cfg` of R-CNN.
Returns:
Tuple[Tensor]: Ground truth for proposals
in a single image. Containing the following Tensors:
- labels(Tensor): Gt_labels for all proposals, has
shape (num_proposals,).
- label_weights(Tensor): Labels_weights for all
proposals, has shape (num_proposals,).
- bbox_targets(Tensor):Regression target for all
proposals, has shape (num_proposals, 4), the
last dimension 4 represents [tl_x, tl_y, br_x, br_y].
- bbox_weights(Tensor):Regression weights for all
proposals, has shape (num_proposals, 4).
"""
num_pos = pos_bboxes.size(0)
num_neg = neg_bboxes.size(0)
num_samples = num_pos + num_neg
# original implementation uses new_zeros since BG are set to be 0
# now use empty & fill because BG cat_id = num_classes,
# FG cat_id = [0, num_classes-1]
labels = pos_bboxes.new_full((num_samples, ),
self.num_classes,
dtype=torch.long)
label_weights = pos_bboxes.new_zeros(num_samples)
bbox_targets = pos_bboxes.new_zeros(num_samples, 4)
bbox_weights = pos_bboxes.new_zeros(num_samples, 4)
if num_pos > 0:
labels[:num_pos] = pos_gt_labels
pos_weight = 1.0 if cfg.pos_weight <= 0 else cfg.pos_weight
label_weights[:num_pos] = pos_weight
if not self.reg_decoded_bbox:
pos_bbox_targets = self.bbox_coder.encode(
pos_bboxes, pos_gt_bboxes)
else:
# When the regression loss (e.g. `IouLoss`, `GIouLoss`)
# is applied directly on the decoded bounding boxes, both
# the predicted boxes and regression targets should be with
# absolute coordinate format.
pos_bbox_targets = pos_gt_bboxes
bbox_targets[:num_pos, :] = pos_bbox_targets
bbox_weights[:num_pos, :] = 1
if num_neg > 0:
label_weights[-num_neg:] = 1.0
return labels, label_weights, bbox_targets, bbox_weights
def get_targets(self,
sampling_results,
gt_bboxes,
gt_labels,
rcnn_train_cfg,
concat=True):
"""Calculate the ground truth for all samples in a batch according to
the sampling_results.
Almost the same as the implementation in bbox_head, we passed
additional parameters pos_inds_list and neg_inds_list to
`_get_target_single` function.
Args:
sampling_results (List[obj:SamplingResults]): Assign results of
all images in a batch after sampling.
gt_bboxes (list[Tensor]): Gt_bboxes of all images in a batch,
each tensor has shape (num_gt, 4), the last dimension 4
represents [tl_x, tl_y, br_x, br_y].
gt_labels (list[Tensor]): Gt_labels of all images in a batch,
each tensor has shape (num_gt,).
rcnn_train_cfg (obj:ConfigDict): `train_cfg` of RCNN.
concat (bool): Whether to concatenate the results of all
the images in a single batch.
Returns:
Tuple[Tensor]: Ground truth for proposals in a single image.
Containing the following list of Tensors:
- labels (list[Tensor],Tensor): Gt_labels for all
proposals in a batch, each tensor in list has
shape (num_proposals,) when `concat=False`, otherwise
just a single tensor has shape (num_all_proposals,).
- label_weights (list[Tensor]): Labels_weights for
all proposals in a batch, each tensor in list has
shape (num_proposals,) when `concat=False`, otherwise
just a single tensor has shape (num_all_proposals,).
- bbox_targets (list[Tensor],Tensor): Regression target
for all proposals in a batch, each tensor in list
has shape (num_proposals, 4) when `concat=False`,
otherwise just a single tensor has shape
(num_all_proposals, 4), the last dimension 4 represents
[tl_x, tl_y, br_x, br_y].
- bbox_weights (list[tensor],Tensor): Regression weights for
all proposals in a batch, each tensor in list has shape
(num_proposals, 4) when `concat=False`, otherwise just a
single tensor has shape (num_all_proposals, 4).
"""
pos_bboxes_list = [res.pos_bboxes for res in sampling_results]
neg_bboxes_list = [res.neg_bboxes for res in sampling_results]
pos_gt_bboxes_list = [res.pos_gt_bboxes for res in sampling_results]
pos_gt_labels_list = [res.pos_gt_labels for res in sampling_results]
labels, label_weights, bbox_targets, bbox_weights = multi_apply(
self._get_target_single,
pos_bboxes_list,
neg_bboxes_list,
pos_gt_bboxes_list,
pos_gt_labels_list,
cfg=rcnn_train_cfg)
if concat:
labels = torch.cat(labels, 0)
label_weights = torch.cat(label_weights, 0)
bbox_targets = torch.cat(bbox_targets, 0)
bbox_weights = torch.cat(bbox_weights, 0)
return labels, label_weights, bbox_targets, bbox_weights
@force_fp32(apply_to=('cls_score', 'bbox_pred'))
def loss(self,
cls_score,
bbox_pred,
rois,
labels,
label_weights,
bbox_targets,
bbox_weights,
reduction_override=None):
losses = dict()
if cls_score is not None:
avg_factor = max(torch.sum(label_weights > 0).float().item(), 1.)
if cls_score.numel() > 0:
losses['loss_cls'] = self.loss_cls(
cls_score,
labels,
label_weights,
avg_factor=avg_factor,
reduction_override=reduction_override)
losses['acc'] = accuracy(cls_score, labels)
if bbox_pred is not None:
bg_class_ind = self.num_classes
# 0~self.num_classes-1 are FG, self.num_classes is BG
pos_inds = (labels >= 0) & (labels < bg_class_ind)
# do not perform bounding box regression for BG anymore.
if pos_inds.any():
if self.reg_decoded_bbox:
# When the regression loss (e.g. `IouLoss`,
# `GIouLoss`, `DIouLoss`) is applied directly on
# the decoded bounding boxes, it decodes the
# already encoded coordinates to absolute format.
bbox_pred = self.bbox_coder.decode(rois[:, 1:], bbox_pred)
if self.reg_class_agnostic:
pos_bbox_pred = bbox_pred.view(
bbox_pred.size(0), 4)[pos_inds.type(torch.bool)]
else:
pos_bbox_pred = bbox_pred.view(
bbox_pred.size(0), -1,
4)[pos_inds.type(torch.bool),
labels[pos_inds.type(torch.bool)]]
losses['loss_bbox'] = self.loss_bbox(
pos_bbox_pred,
bbox_targets[pos_inds.type(torch.bool)],
bbox_weights[pos_inds.type(torch.bool)],
avg_factor=bbox_targets.size(0),
reduction_override=reduction_override)
else:
losses['loss_bbox'] = bbox_pred[pos_inds].sum()
return losses
@force_fp32(apply_to=('cls_score', 'bbox_pred'))
def get_bboxes(self,
rois,
cls_score,
bbox_pred,
img_shape,
scale_factor,
rescale=False,
cfg=None):
"""Transform network output for a batch into bbox predictions.
If the input rois has batch dimension, the function would be in
`batch_mode` and return is a tuple[list[Tensor], list[Tensor]],
otherwise, the return is a tuple[Tensor, Tensor].
Args:
rois (Tensor): Boxes to be transformed. Has shape (num_boxes, 5)
or (B, num_boxes, 5)
cls_score (list[Tensor] or Tensor): Box scores for
each scale level, each is a 4D-tensor, the channel number is
num_points * num_classes.
bbox_pred (Tensor, optional): Box energies / deltas for each scale
level, each is a 4D-tensor, the channel number is
num_classes * 4.
img_shape (Sequence[int] or torch.Tensor or Sequence[
Sequence[int]], optional): Maximum bounds for boxes, specifies
(H, W, C) or (H, W). If rois shape is (B, num_boxes, 4), then
the max_shape should be a Sequence[Sequence[int]]
and the length of max_shape should also be B.
scale_factor (tuple[ndarray] or ndarray): Scale factor of the
image arange as (w_scale, h_scale, w_scale, h_scale). In
`batch_mode`, the scale_factor shape is tuple[ndarray].
rescale (bool): If True, return boxes in original image space.
Default: False.
cfg (obj:`ConfigDict`): `test_cfg` of Bbox Head. Default: None
Returns:
tuple[list[Tensor], list[Tensor]] or tuple[Tensor, Tensor]:
If the input has a batch dimension, the return value is
a tuple of the list. The first list contains the boxes of
the corresponding image in a batch, each tensor has the
shape (num_boxes, 5) and last dimension 5 represent
(tl_x, tl_y, br_x, br_y, score). Each Tensor in the second
list is the labels with shape (num_boxes, ). The length of
both lists should be equal to batch_size. Otherwise return
value is a tuple of two tensors, the first tensor is the
boxes with scores, the second tensor is the labels, both
have the same shape as the first case.
"""
if isinstance(cls_score, list):
cls_score = sum(cls_score) / float(len(cls_score))
scores = F.softmax(
cls_score, dim=-1) if cls_score is not None else None
batch_mode = True
if rois.ndim == 2:
# e.g. AugTest, Cascade R-CNN, HTC, SCNet...
batch_mode = False
# add batch dimension
if scores is not None:
scores = scores.unsqueeze(0)
if bbox_pred is not None:
bbox_pred = bbox_pred.unsqueeze(0)
rois = rois.unsqueeze(0)
if bbox_pred is not None:
bboxes = self.bbox_coder.decode(
rois[..., 1:], bbox_pred, max_shape=img_shape)
else:
bboxes = rois[..., 1:].clone()
if img_shape is not None:
max_shape = bboxes.new_tensor(img_shape)[..., :2]
min_xy = bboxes.new_tensor(0)
max_xy = torch.cat(
[max_shape] * 2, dim=-1).flip(-1).unsqueeze(-2)
bboxes = torch.where(bboxes < min_xy, min_xy, bboxes)
bboxes = torch.where(bboxes > max_xy, max_xy, bboxes)
if rescale and bboxes.size(-2) > 0:
if not isinstance(scale_factor, tuple):
scale_factor = tuple([scale_factor])
# B, 1, bboxes.size(-1)
scale_factor = bboxes.new_tensor(scale_factor).unsqueeze(1).repeat(
1, 1,
bboxes.size(-1) // 4)
bboxes /= scale_factor
det_bboxes = []
det_labels = []
for (bbox, score) in zip(bboxes, scores):
if cfg is not None:
det_bbox, det_label = multiclass_nms(bbox, score,
cfg.score_thr, cfg.nms,
cfg.max_per_img)
else:
det_bbox, det_label = bbox, score
det_bboxes.append(det_bbox)
det_labels.append(det_label)
if not batch_mode:
det_bboxes = det_bboxes[0]
det_labels = det_labels[0]
return det_bboxes, det_labels
@force_fp32(apply_to=('bbox_preds', ))
def refine_bboxes(self, rois, labels, bbox_preds, pos_is_gts, img_metas):
"""Refine bboxes during training.
Args:
rois (Tensor): Shape (n*bs, 5), where n is image number per GPU,
and bs is the sampled RoIs per image. The first column is
the image id and the next 4 columns are x1, y1, x2, y2.
labels (Tensor): Shape (n*bs, ).
bbox_preds (Tensor): Shape (n*bs, 4) or (n*bs, 4*#class).
pos_is_gts (list[Tensor]): Flags indicating if each positive bbox
is a gt bbox.
img_metas (list[dict]): Meta info of each image.
Returns:
list[Tensor]: Refined bboxes of each image in a mini-batch.
Example:
>>> # xdoctest: +REQUIRES(module:kwarray)
>>> import kwarray
>>> import numpy as np
>>> from mmdet.core.bbox.demodata import random_boxes
>>> self = BBoxHead(reg_class_agnostic=True)
>>> n_roi = 2
>>> n_img = 4
>>> scale = 512
>>> rng = np.random.RandomState(0)
>>> img_metas = [{'img_shape': (scale, scale)}
... for _ in range(n_img)]
>>> # Create rois in the expected format
>>> roi_boxes = random_boxes(n_roi, scale=scale, rng=rng)
>>> img_ids = torch.randint(0, n_img, (n_roi,))
>>> img_ids = img_ids.float()
>>> rois = torch.cat([img_ids[:, None], roi_boxes], dim=1)
>>> # Create other args
>>> labels = torch.randint(0, 2, (n_roi,)).long()
>>> bbox_preds = random_boxes(n_roi, scale=scale, rng=rng)
>>> # For each image, pretend random positive boxes are gts
>>> is_label_pos = (labels.numpy() > 0).astype(np.int)
>>> lbl_per_img = kwarray.group_items(is_label_pos,
... img_ids.numpy())
>>> pos_per_img = [sum(lbl_per_img.get(gid, []))
... for gid in range(n_img)]
>>> pos_is_gts = [
>>> torch.randint(0, 2, (npos,)).byte().sort(
>>> descending=True)[0]
>>> for npos in pos_per_img
>>> ]
>>> bboxes_list = self.refine_bboxes(rois, labels, bbox_preds,
>>> pos_is_gts, img_metas)
>>> print(bboxes_list)
"""
img_ids = rois[:, 0].long().unique(sorted=True)
assert img_ids.numel() <= len(img_metas)
bboxes_list = []
for i in range(len(img_metas)):
inds = torch.nonzero(
rois[:, 0] == i, as_tuple=False).squeeze(dim=1)
num_rois = inds.numel()
bboxes_ = rois[inds, 1:]
label_ = labels[inds]
bbox_pred_ = bbox_preds[inds]
img_meta_ = img_metas[i]
pos_is_gts_ = pos_is_gts[i]
bboxes = self.regress_by_class(bboxes_, label_, bbox_pred_,
img_meta_)
# filter gt bboxes
pos_keep = 1 - pos_is_gts_
keep_inds = pos_is_gts_.new_ones(num_rois)
keep_inds[:len(pos_is_gts_)] = pos_keep
bboxes_list.append(bboxes[keep_inds.type(torch.bool)])
return bboxes_list
@force_fp32(apply_to=('bbox_pred', ))
def regress_by_class(self, rois, label, bbox_pred, img_meta):
"""Regress the bbox for the predicted class. Used in Cascade R-CNN.
Args:
rois (Tensor): shape (n, 4) or (n, 5)
label (Tensor): shape (n, )
bbox_pred (Tensor): shape (n, 4*(#class)) or (n, 4)
img_meta (dict): Image meta info.
Returns:
Tensor: Regressed bboxes, the same shape as input rois.
"""
assert rois.size(1) == 4 or rois.size(1) == 5, repr(rois.shape)
if not self.reg_class_agnostic:
label = label * 4
inds = torch.stack((label, label + 1, label + 2, label + 3), 1)
bbox_pred = torch.gather(bbox_pred, 1, inds)
assert bbox_pred.size(1) == 4
if rois.size(1) == 4:
new_rois = self.bbox_coder.decode(
rois, bbox_pred, max_shape=img_meta['img_shape'])
else:
bboxes = self.bbox_coder.decode(
rois[:, 1:], bbox_pred, max_shape=img_meta['img_shape'])
new_rois = torch.cat((rois[:, [0]], bboxes), dim=1)
return new_rois
|