-
Notifications
You must be signed in to change notification settings - Fork 0
/
trainer.py
413 lines (339 loc) · 17.5 KB
/
trainer.py
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
import torch
from models.backbone import resnet50
from models.rpn import RegionProposalNetwork
from models.RoIHead import Resnet50RoIHead
from torch import nn, optim
from tqdm import tqdm
import numpy as np
import torch.nn.functional as F
from utils.box_decoder import DecodeBox
from utils.utils import get_new_img_size, resize_image
from utils.utils_map import get_map
from PIL import Image
import os
import warnings
warnings.filterwarnings("ignore")
def bbox_iou(bbox_a, bbox_b):
if bbox_a.shape[1] != 4 or bbox_b.shape[1] != 4:
print(bbox_a, bbox_b)
raise IndexError
tl = np.maximum(bbox_a[:, None, :2], bbox_b[:, :2])
br = np.minimum(bbox_a[:, None, 2:], bbox_b[:, 2:])
area_i = np.prod(br - tl, axis=2) * (tl < br).all(axis=2)
area_a = np.prod(bbox_a[:, 2:] - bbox_a[:, :2], axis=1)
area_b = np.prod(bbox_b[:, 2:] - bbox_b[:, :2], axis=1)
return area_i / (area_a[:, None] + area_b - area_i)
def bbox2loc(src_bbox, dst_bbox):
width = src_bbox[:, 2] - src_bbox[:, 0]
height = src_bbox[:, 3] - src_bbox[:, 1]
ctr_x = src_bbox[:, 0] + 0.5 * width
ctr_y = src_bbox[:, 1] + 0.5 * height
base_width = dst_bbox[:, 2] - dst_bbox[:, 0]
base_height = dst_bbox[:, 3] - dst_bbox[:, 1]
base_ctr_x = dst_bbox[:, 0] + 0.5 * base_width
base_ctr_y = dst_bbox[:, 1] + 0.5 * base_height
eps = np.finfo(height.dtype).eps
width = np.maximum(width, eps)
height = np.maximum(height, eps)
dx = (base_ctr_x - ctr_x) / width
dy = (base_ctr_y - ctr_y) / height
dw = np.log(base_width / width)
dh = np.log(base_height / height)
loc = np.vstack((dx, dy, dw, dh)).transpose()
return loc
class AnchorGenerator(object):
def __init__(self, n_sample=256, pos_iou_thresh=0.7, neg_iou_thresh=0.3, pos_ratio=0.5):
self.n_sample = n_sample
self.pos_iou_thresh = pos_iou_thresh
self.neg_iou_thresh = neg_iou_thresh
self.pos_ratio = pos_ratio
def calc_ious(self, anchor, bbox):
ious = bbox_iou(anchor, bbox)
if len(bbox) == 0:
return np.zeros(len(anchor), np.int32), np.zeros(len(anchor)), np.zeros(len(bbox))
argmax_ious = ious.argmax(axis=1)
max_ious = np.max(ious, axis=1)
gt_argmax_ious = ious.argmax(axis=0)
for i in range(len(gt_argmax_ious)):
argmax_ious[gt_argmax_ious[i]] = i
return argmax_ious, max_ious, gt_argmax_ious
def create_label(self, anchor, bbox):
label = np.empty((len(anchor),), dtype=np.int32)
label.fill(-1)
argmax_ious, max_ious, gt_argmax_ious = self.calc_ious(anchor, bbox)
label[max_ious < self.neg_iou_thresh] = 0
label[max_ious >= self.pos_iou_thresh] = 1
if len(gt_argmax_ious) > 0:
label[gt_argmax_ious] = 1
n_pos = int(self.pos_ratio * self.n_sample)
pos_index = np.where(label == 1)[0]
if len(pos_index) > n_pos:
disable_index = np.random.choice(pos_index, size=(len(pos_index) - n_pos), replace=False)
label[disable_index] = -1
n_neg = self.n_sample - np.sum(label == 1)
neg_index = np.where(label == 0)[0]
if len(neg_index) > n_neg:
disable_index = np.random.choice(neg_index, size=(len(neg_index) - n_neg), replace=False)
label[disable_index] = -1
return argmax_ious, label
def __call__(self, bbox, anchor):
argmax_ious, label = self.create_label(anchor, bbox)
if (label > 0).any():
loc = bbox2loc(anchor, bbox[argmax_ious])
return loc, label
else:
return np.zeros_like(anchor), label
class Proposal(object):
def __init__(self, n_sample=128, pos_ratio=0.5, pos_iou_thresh=0.5, neg_iou_thresh_high=0.5, neg_iou_thresh_low=0):
self.n_sample = n_sample
self.pos_ratio = pos_ratio
self.pos_roi_per_image = np.round(self.n_sample * self.pos_ratio)
self.pos_iou_thresh = pos_iou_thresh
self.neg_iou_thresh_high = neg_iou_thresh_high
self.neg_iou_thresh_low = neg_iou_thresh_low
def __call__(self, roi, bbox, label, loc_normalize_std=(0.1, 0.1, 0.2, 0.2)):
roi = np.concatenate((roi.detach().cpu().numpy(), bbox), axis=0)
iou = bbox_iou(roi, bbox)
if len(bbox) == 0:
gt_assignment = np.zeros(len(roi), np.int32)
max_iou = np.zeros(len(roi))
gt_roi_label = np.zeros(len(roi))
else:
gt_assignment = iou.argmax(axis=1)
max_iou = iou.max(axis=1)
gt_roi_label = label[gt_assignment] + 1
pos_index = np.where(max_iou >= self.pos_iou_thresh)[0]
pos_roi_per_this_image = int(min(self.pos_roi_per_image, pos_index.size))
if pos_index.size > 0:
pos_index = np.random.choice(pos_index, size=pos_roi_per_this_image, replace=False)
neg_index = np.where((max_iou < self.neg_iou_thresh_high) & (max_iou >= self.neg_iou_thresh_low))[0]
neg_roi_per_this_image = self.n_sample - pos_roi_per_this_image
neg_roi_per_this_image = int(min(neg_roi_per_this_image, neg_index.size))
if neg_index.size > 0:
neg_index = np.random.choice(neg_index, size=neg_roi_per_this_image, replace=False)
keep_index = np.append(pos_index, neg_index)
sample_roi = roi[keep_index]
if len(bbox) == 0:
return sample_roi, np.zeros_like(sample_roi), gt_roi_label[keep_index]
gt_roi_loc = bbox2loc(sample_roi, bbox[gt_assignment[keep_index]])
gt_roi_loc = (gt_roi_loc / np.array(loc_normalize_std, np.float32))
gt_roi_label = gt_roi_label[keep_index]
gt_roi_label[pos_roi_per_this_image:] = 0
return sample_roi, gt_roi_loc, gt_roi_label
class FRCNNTrainer(object):
def __init__(self, dataloader, device, opt, num_classes, class_names):
self.opt = opt
self.device = device
self.dataloader = dataloader
self.epochs = opt.epochs
self.rpn_sigma = 1.0
self.roi_sigma = 1.0
self.loc_normalize_std = [0.1, 0.1, 0.2, 0.2]
self.epoch = [0]
self.map = [0]
self.best_map = 0
self.nms_iou = opt.nms_iou
self.confidence = opt.confidence
self.max_boxes = opt.max_boxes
self.map_out_path = opt.map_out_path
self.class_names = class_names
self.MINOVERLAP = opt.MINOVERLAP
if not os.path.exists(self.map_out_path):
os.mkdir(self.map_out_path)
if not os.path.exists(os.path.join(self.map_out_path, "detection-results")):
os.mkdir(os.path.join(self.map_out_path, "detection-results"))
if not os.path.exists(os.path.join(self.map_out_path, "ground-truth")):
os.mkdir(os.path.join(self.map_out_path, "ground-truth"))
self.extractor, self.classifier = resnet50(opt.pretrain_backbone)
self.extractor = self.extractor.to(self.device)
self.rpn = RegionProposalNetwork(
1024, 512,
ratios=opt.ratios,
anchor_scales=opt.scales,
feat_stride=16,
mode="training"
).to(device)
self.anchor_generator = AnchorGenerator()
self.proposal = Proposal()
self.head = Resnet50RoIHead(
n_class=num_classes + 1,
roi_size=38,
spatial_scale=1,
classifier=self.classifier
).to(device)
self.optimizer = optim.Adam([{"params": self.extractor.parameters()},
{"params": self.rpn.parameters()},
{"params": self.head.parameters()}],
lr=self.opt.lr)
self.std = torch.Tensor([0.1, 0.1, 0.2, 0.2]).repeat(num_classes + 1)[None]
self.decoder = DecodeBox(self.std.to(self.device), num_classes)
self.flag = False
if opt.freeze_backbone:
for name, param in self.extractor.named_parameters():
param.requires_grad_(False)
self.flag = True
print("backbone parameters have been frozen!")
def fast_rcnn_loc_loss(self, pred_loc, gt_loc, gt_label, sigma):
pred_loc = pred_loc[gt_label > 0]
gt_loc = gt_loc[gt_label > 0]
sigma_squared = sigma ** 2
regression_diff = (gt_loc - pred_loc)
regression_diff = regression_diff.abs().float()
regression_loss = torch.where(
regression_diff < (1. / sigma_squared),
0.5 * sigma_squared * regression_diff ** 2,
regression_diff - 0.5 / sigma_squared
)
regression_loss = regression_loss.sum()
num_pos = (gt_label > 0).sum().float()
regression_loss /= torch.max(num_pos, torch.ones_like(num_pos))
return regression_loss
def train_step(self, epoch):
if self.opt.freeze_backbone and self.flag and epoch >= self.opt.freeze_epoch:
for name, param in self.extractor.named_parameters():
param.requires_grad_(True)
self.flag = False
print("backbone parameters have been unfrozen!")
self.extractor.train()
self.rpn.train()
self.head.train()
with tqdm(total=len(self.dataloader), desc=f'Epoch {epoch + 1}/{self.epochs}', postfix=dict, mininterval=0.3) as pbar:
for k, batch in enumerate(self.dataloader):
self.optimizer.zero_grad()
images, boxes, labels = batch[0], batch[1], batch[2]
img_size = images.shape[-2:]
n = images.shape[0]
images = images.to(self.device)
base_features = self.extractor(images)
rpn_locs, rpn_scores, rois, roi_indices, anchors = self.rpn(base_features, img_size, self.opt.scale)
rpn_loc_loss_all, rpn_cls_loss_all, roi_loc_loss_all, roi_cls_loss_all = 0, 0, 0, 0
sample_rois, sample_indexes, gt_roi_locs, gt_roi_labels = [], [], [], []
for i in range(n):
bbox = boxes[i]
label = labels[i]
rpn_loc = rpn_locs[i]
rpn_score = rpn_scores[i]
roi = rois[i]
gt_rpn_loc, gt_rpn_label = self.anchor_generator(bbox, anchors[0].cpu().numpy())
gt_rpn_loc = torch.Tensor(gt_rpn_loc).type_as(rpn_locs)
gt_rpn_label = torch.Tensor(gt_rpn_label).type_as(rpn_locs).long()
rpn_loc_loss = self.fast_rcnn_loc_loss(rpn_loc, gt_rpn_loc, gt_rpn_label, self.rpn_sigma)
rpn_cls_loss = F.cross_entropy(rpn_score, gt_rpn_label, ignore_index=-1)
rpn_loc_loss_all += rpn_loc_loss
rpn_cls_loss_all += rpn_cls_loss
sample_roi, gt_roi_loc, gt_roi_label = self.proposal(roi, bbox, label, self.loc_normalize_std)
sample_rois.append(torch.Tensor(sample_roi).type_as(rpn_locs))
sample_indexes.append(torch.ones(len(sample_roi)).type_as(rpn_locs) * roi_indices[i][0])
gt_roi_locs.append(torch.Tensor(gt_roi_loc).type_as(rpn_locs))
gt_roi_labels.append(torch.Tensor(gt_roi_label).type_as(rpn_locs).long())
sample_rois = torch.stack(sample_rois, dim=0)
sample_indexes = torch.stack(sample_indexes, dim=0)
roi_cls_locs, roi_scores = self.head(base_features, sample_rois, sample_indexes, img_size, self.device)
for i in range(n):
n_sample = roi_cls_locs.shape[1]
roi_cls_loc = roi_cls_locs[i]
roi_score = roi_scores[i]
gt_roi_loc = gt_roi_locs[i]
gt_roi_label = gt_roi_labels[i]
roi_cls_loc = roi_cls_loc.view(n_sample, -1, 4)
roi_loc = roi_cls_loc[torch.arange(0, n_sample), gt_roi_label]
roi_loc_loss = self.fast_rcnn_loc_loss(roi_loc, gt_roi_loc, gt_roi_label.data, self.roi_sigma)
roi_cls_loss = nn.CrossEntropyLoss()(roi_score, gt_roi_label)
roi_loc_loss_all += roi_loc_loss
roi_cls_loss_all += roi_cls_loss
losses = [rpn_loc_loss_all / n, rpn_cls_loss_all / n, roi_loc_loss_all / n, roi_cls_loss_all / n]
losses = losses + [sum(losses)]
losses[-1].backward()
self.optimizer.step()
pbar.set_postfix(**{"total_loss": losses[-1].item(),
"rpn_loc_loss": losses[0].item(),
"rpn_cls_loss": losses[1].item(),
"roi_loc_loss": losses[2].item(),
"roi_cls_loss": losses[3].item()})
pbar.update(1)
def generate_txt(self, valid_lines, epoch):
print("Start generate result file....")
self.extractor.eval()
self.rpn.eval()
self.head.eval()
with tqdm(total=len(valid_lines), desc=f'Epoch {epoch + 1}/{self.epochs}', postfix=dict,
mininterval=0.3) as pbar:
for annotation_line in valid_lines:
line = annotation_line.split()
image_id = os.path.basename(line[0]).split('.')[0]
f = open(os.path.join(self.map_out_path, "detection-results/" + image_id + ".txt"), "w")
image = Image.open(line[0])
gt_boxes = np.array([np.array(list(map(int, box.split(',')))) for box in line[1:]])
if epoch == 0:
with open(os.path.join(self.map_out_path, "ground-truth/" + image_id + ".txt"), "w") as new_f:
for box in gt_boxes:
left, top, right, bottom, obj = box
obj_name = self.class_names[obj]
new_f.write("%s %s %s %s %s\n" % (obj_name, left, top, right, bottom))
image_shape = np.array(np.shape(image)[0:2])
input_shape = get_new_img_size(image_shape[0], image_shape[1])
image_data = np.array(resize_image(image, [input_shape[1], input_shape[0]]), dtype=np.float32) / 255.0
image_data = np.transpose(image_data, (2, 0, 1))[None]
with torch.no_grad():
images = torch.from_numpy(image_data).to(self.device)
roi_cls_locs, roi_scores, rois, _ = self.predict(images)
result = self.decoder.forward(roi_cls_locs, roi_scores, rois,
image_shape, input_shape,
nms_iou=self.nms_iou, confidence=self.confidence)
if len(result[0]) <= 0:
f.close()
continue
top_label = np.array(result[0][:, 5], dtype='int32')
top_conf = result[0][:, 4]
top_boxes = result[0][:, :4]
top_100 = np.argsort(top_conf)[::-1][:self.max_boxes]
top_boxes = top_boxes[top_100]
top_conf = top_conf[top_100]
top_label = top_label[top_100]
for i, c in list(enumerate(top_label)):
predicted_class = self.class_names[int(c)]
box = top_boxes[i]
score = str(top_conf[i])
top, left, bottom, right = box
if predicted_class not in self.class_names:
continue
f.write("%s %s %s %s %s %s\n" % (
predicted_class, score[:6], str(int(left)), str(int(top)), str(int(right)), str(int(bottom))))
f.close()
pbar.set_postfix(**{"file": image_id})
pbar.update(1)
print("result files have been generated sucessfully")
def calc_map(self):
print("Start calculate mAP")
mAP = get_map(self.MINOVERLAP, self.opt.draw_plot, path=self.map_out_path)
print("mAP has been calculated")
return mAP
def predict(self, images):
img_size = images.shape[-2:]
images = images.to(self.device)
base_feature = self.extractor(images)
rpn_locs, rpn_scores, rois, roi_indices, anchors = self.rpn(base_feature, img_size, self.opt.scale)
roi_cls_locs, roi_scores = self.head(base_feature, rois, roi_indices, img_size, self.device)
return roi_cls_locs, roi_scores, rois, roi_indices
def save(self, epoch, best=False):
ckpt = {
"backbone": self.extractor.state_dict(),
"rpn": self.rpn.state_dict(),
"head": self.head.state_dict(),
"epoch": epoch,
"best_map": self.best_map,
"maps": self.map,
"epochs": self.epoch
}
torch.save(ckpt, self.opt.save_dir + "/last.pth")
if best:
torch.save(ckpt, self.opt.save_dir + "/best.pth")
def load(self, path):
ckpt = torch.load(path)
self.extractor.load_state_dict(ckpt["backbone"])
self.rpn.load_state_dict(ckpt["rpn"])
self.head.load_state_dict(ckpt["head"])
self.epoch = ckpt["epochs"]
self.map = ckpt["maps"]
self.best_map = ckpt["best_map"]
return ckpt["epoch"]