-
Notifications
You must be signed in to change notification settings - Fork 1
/
engine_vg.py
291 lines (256 loc) · 14 KB
/
engine_vg.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
"""
Train and eval functions used in main.py
"""
import math
import os
import sys
import json
from typing import Iterable
import torch
import util.misc as utils
from PIL import Image, ImageDraw
import matplotlib.pyplot as plt
import numpy as np
import torch.nn.functional as F
# from datasets.data_prefetcher import data_prefetcher
# # Reuse Deformable DETR's train function
# from engine import train_one_epoch
def train_one_epoch(model: torch.nn.Module, criterion: torch.nn.Module,
data_loader: Iterable, optimizer: torch.optim.Optimizer,
lr_scheduler: torch.optim.lr_scheduler._LRScheduler,
device: torch.device, epoch: int, max_norm: float = 0):
model.train()
criterion.train()
metric_logger = utils.MetricLogger(delimiter=" ")
metric_logger.add_meter('lr', utils.SmoothedValue(window_size=1, fmt='{value:.6f}'))
# metric_logger.add_meter('class_error', utils.SmoothedValue(window_size=1, fmt='{value:.2f}'))
metric_logger.add_meter('grad_norm', utils.SmoothedValue(window_size=1, fmt='{value:.2f}'))
header = 'Epoch: [{}]'.format(epoch)
print_freq = 50
prefetcher = data_prefetcher(data_loader, device, prefetch=True)
samples, targets = prefetcher.next()
# for samples, targets in metric_logger.log_every(data_loader, print_freq, header):
for _ in metric_logger.log_every(range(len(data_loader)), print_freq, header):
outputs = model(samples)
loss_dict = criterion(outputs, targets)
weight_dict = criterion.weight_dict
losses = sum(loss_dict[k] * weight_dict[k] for k in loss_dict.keys() if k in weight_dict)
# reduce losses over all GPUs for logging purposes
loss_dict_reduced = utils.reduce_dict(loss_dict)
loss_dict_reduced_unscaled = {f'{k}_unscaled': v
for k, v in loss_dict_reduced.items()}
loss_dict_reduced_scaled = {k: v * weight_dict[k]
for k, v in loss_dict_reduced.items() if k in weight_dict}
losses_reduced_scaled = sum(loss_dict_reduced_scaled.values())
loss_value = losses_reduced_scaled.item()
if not math.isfinite(loss_value):
print("Loss is {}, stopping training".format(loss_value))
print(loss_dict_reduced)
sys.exit(1)
optimizer.zero_grad()
losses.backward()
if max_norm > 0:
grad_total_norm = torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm)
else:
grad_total_norm = utils.get_total_grad_norm(model.parameters(), max_norm)
optimizer.step()
lr_scheduler.step()
metric_logger.update(loss=loss_value, **loss_dict_reduced_scaled, **loss_dict_reduced_unscaled)
# metric_logger.update(class_error=loss_dict_reduced['class_error'])
metric_logger.update(lr=optimizer.param_groups[0]["lr"])
metric_logger.update(grad_norm=grad_total_norm)
samples, targets = prefetcher.next()
# gather the stats from all processes
metric_logger.synchronize_between_processes()
print("Averaged stats:", metric_logger)
return {k: meter.global_avg for k, meter in metric_logger.meters.items()}
from util.box_ops import box_iou, box_cxcywh_to_xyxy, mask_iou
@torch.no_grad()
def evaluate(model, criterion, postprocessors, data_loader, device, output_dir, visualize=False):
model.eval()
criterion.eval()
# visualize=False
if visualize:
split_name = data_loader.dataset.split
output_dir = output_dir / 'vis' / split_name
output_dir.mkdir(parents=True, exist_ok=True)
(output_dir / 'mask').mkdir(parents=True, exist_ok=True)
(output_dir / 'bbox').mkdir(parents=True, exist_ok=True)
(output_dir / 'att').mkdir(parents=True, exist_ok=True)
(output_dir / 'gt').mkdir(parents=True, exist_ok=True)
purple = np.array([[[128, 0, 128]]], dtype=np.uint8)
yellow = np.array([[[255, 255, 0]]], dtype=np.uint8)
metric_logger = utils.MetricLogger(delimiter=" ")
# metric_logger.add_meter('class_error', utils.SmoothedValue(window_size=1, fmt='{value:.2f}'))
header = data_loader.dataset.split + ':'
results_dict = {}
results_iou = {'det':{}, 'seg':{}}
sum_accu = 0.
sum_iou = 0.
cnt_test = 0.
seg_iou = 0.
cnt_seg = 0.
for samples, targets in metric_logger.log_every(data_loader, 50, header):
samples = {k: v.to(device, non_blocking=True) for k, v in samples.items()}
targets = [{k: v.to(device) for k, v in t.items()} for t in targets]
outputs = model(samples)
loss_dict = criterion(outputs, targets)
weight_dict = criterion.weight_dict
# reduce losses over all GPUs for logging purposes
loss_dict_reduced = utils.reduce_dict(loss_dict)
loss_dict_reduced_scaled = {k: v * weight_dict[k]
for k, v in loss_dict_reduced.items() if k in weight_dict}
loss_dict_reduced_unscaled = {f'{k}_unscaled': v
for k, v in loss_dict_reduced.items()}
metric_logger.update(loss=sum(loss_dict_reduced_scaled.values()),
**loss_dict_reduced_scaled,
**loss_dict_reduced_unscaled)
# metric_logger.update(class_error=loss_dict_reduced['class_error'])
# TODO： some issues with data loaders here
orig_target_sizes = torch.stack([t["orig_size"] for t in targets], dim=0)
results = postprocessors['bbox'](outputs, orig_target_sizes)
# res = {target['image_id'].item(): output for target, output in zip(targets, results)}
for i, res in enumerate(results):
t = box_cxcywh_to_xyxy(targets[i]['boxes'])
assert t.size(0) == res['boxes'].size(0), (res, t)
iou, union = box_iou(t, res['boxes'])
iou = torch.diag(iou)
# print(t, res['boxes'], iou, union)
sum_accu = sum_accu + torch.sum((iou > 0.5).type(torch.float))#.item()
sum_iou = sum_iou + torch.sum(iou)#.item()
cnt_test = cnt_test + torch.tensor(len(targets[i]['boxes']), device=sum_iou.device)
results_iou['det'][targets[i]['dataset_id'].item()] = torch.sum(iou).item()
results_scaled = postprocessors['bbox'](outputs, orig_target_sizes, scale_to_original_shape=True)
# TODO support multi-phrase in the future
if 'segm' in postprocessors.keys():
target_sizes = torch.stack([t["size"] for t in targets], dim=0)
results = postprocessors['segm'](results, outputs, orig_target_sizes, target_sizes)
for i, res in enumerate(results):
t = targets[i]
t_mask = t['masks']
pred_mask = res['masks'][0]
# print(pred_mask.shape, t_mask.shape)
iou = mask_iou(pred_mask[0], t_mask)
seg_iou = seg_iou + iou
cnt_seg = cnt_seg + 1
results_iou['seg'][targets[i]['dataset_id'].item()] = iou.item()
if visualize:
dataset_id = t['dataset_id'].item()
pred_mask = res['masks_origin'][0, 0].cpu().unsqueeze(-1).numpy().astype(np.uint8)
img, mask, phrase, tgt_box, img_file = data_loader.dataset.pull_item(dataset_id)
assert pred_mask.shape[:2] == mask.shape, (pred_mask.shape, mask.shape)
# print(pred_mask.shape, yellow.shape)
img_name = img_file.split('/')[-1].split('.')[0]
pred_mask = pred_mask * yellow + (1-pred_mask)*purple
# print(pred_mask.shape, yellow.shape)
pred_mask = Image.fromarray(pred_mask)
pred_mask.save(output_dir / 'mask'/ f"{img_name}_{dataset_id:05d}.jpg")
mask = np.expand_dims(mask, -1)
gt = mask * yellow + (1-mask)*purple
# print(pred_mask.shape, yellow.shape)
gt_mask = Image.fromarray(gt)
gt_mask.save(output_dir / 'gt'/ f"{img_name}_{dataset_id:05d}.jpg")
pred_box = results_scaled[i]['boxes'][0].cpu().numpy().tolist()
# print(pred_box, tgt_box)
img_bbox = Image.fromarray(img)
draw = ImageDraw.Draw(img_bbox)
draw.rectangle(pred_box, outline='blue', width=5)
draw.rectangle(tgt_box.tolist(), outline='red', width=5)
img_bbox.save(output_dir / 'bbox'/ f"{img_name}_{dataset_id:05d}.jpg")
att_mask = outputs['mask_att'][i:i+1, :].cpu()
h, w, _ = mask.shape
att_mask = F.interpolate(att_mask, size=(320, 320), mode="bilinear").numpy()
# print(att_mask.shape)
plt.imsave(output_dir / 'att' /f"{img_name}_{dataset_id:05d}_0.jpg", att_mask[0, 0, :h//2, :w//2], cmap='viridis')
plt.imsave(output_dir / 'att' /f"{img_name}_{dataset_id:05d}_1.jpg", att_mask[0, 1, :h//2, :w//2], cmap='viridis')
plt.imsave(output_dir / 'att' /f"{img_name}_{dataset_id:05d}_2.jpg", att_mask[0, 2, :h//2, :w//2], cmap='viridis')
plt.imsave(output_dir / 'att' /f"{img_name}_{dataset_id:05d}_7.jpg", att_mask[0, 7, :h//2, :w//2], cmap='viridis')
# att_mask = att_mask[0, 0, :h, :w, None]
# att_mask_rescaled = (att_mask - att_mask.min()) / (att_mask.max()-att_mask.min())
# att_mask_rescaled = np.clip(1.5 * att_mask_rescaled - 0.5, 0., 1.0)
# att_img = (img * att_mask_rescaled).astype(np.uint8)
# att_img = Image.fromarray(att_img)
# att_img.save(output_dir / 'att' / f"{img_name}_{dataset_id:05d}_0.jpg")
# plt.imsave(output_dir / 'att' /f"0_{img_name}_{dataset_id:05d}.jpg", att_mask[:,:,0], cmap='viridis')
results_dict.update({target['image_id'].item(): output['boxes'].cpu().numpy().tolist() for target, output in zip(targets, results_scaled)})
# gather the stats from all processes
metric_logger.synchronize_between_processes()
print("Averaged stats:", metric_logger)
stats = {k: meter.global_avg for k, meter in metric_logger.meters.items()}
if utils.is_dist_avail_and_initialized():
torch.distributed.all_reduce(sum_accu)
torch.distributed.all_reduce(cnt_test)
torch.distributed.all_reduce(sum_iou)
stats["accuracy_iou0.5"] = (sum_accu / cnt_test).cpu().item()
stats["miou"] = (sum_iou / cnt_test).cpu().item()
if 'segm' in postprocessors.keys():
if utils.is_dist_avail_and_initialized():
torch.distributed.all_reduce(seg_iou)
cnt_seg = utils.get_world_size() * cnt_seg
print(cnt_seg)
stats["seg_miou"] = (seg_iou / cnt_seg).cpu().item()
# do not print aux test loss
stats = {k:v for k,v in stats.items() if k.split('_')[-1] not in ['unscaled', '0', '1', '2']}
# with (output_dir / f"{data_loader.dataset.split}_iou.json").open("w") as f:
# f.write(json.dumps(results_iou) + "\n")
return stats, results_dict
def to_cuda(samples, targets, device):
# samples = samples.to(device, non_blocking=True)
samples = {k: v.to(device, non_blocking=True) for k, v in samples.items()}
targets = [{k: v.to(device, non_blocking=True) for k, v in t.items()} for t in targets]
return samples, targets
class data_prefetcher():
def __init__(self, loader, device, prefetch=True):
self.loader = iter(loader)
self.prefetch = prefetch
self.device = device
if prefetch:
self.stream = torch.cuda.Stream()
self.preload()
def preload(self):
try:
self.next_samples, self.next_targets = next(self.loader)
except StopIteration:
self.next_samples = None
self.next_targets = None
return
# if record_stream() doesn't work, another option is to make sure device inputs are created
# on the main stream.
# self.next_input_gpu = torch.empty_like(self.next_input, device='cuda')
# self.next_target_gpu = torch.empty_like(self.next_target, device='cuda')
# Need to make sure the memory allocated for next_* is not still in use by the main stream
# at the time we start copying to next_*:
# self.stream.wait_stream(torch.cuda.current_stream())
with torch.cuda.stream(self.stream):
self.next_samples, self.next_targets = to_cuda(self.next_samples, self.next_targets, self.device)
# more code for the alternative if record_stream() doesn't work:
# copy_ will record the use of the pinned source tensor in this side stream.
# self.next_input_gpu.copy_(self.next_input, non_blocking=True)
# self.next_target_gpu.copy_(self.next_target, non_blocking=True)
# self.next_input = self.next_input_gpu
# self.next_target = self.next_target_gpu
# With Amp, it isn't necessary to manually convert data to half.
# if args.fp16:
# self.next_input = self.next_input.half()
# else:
def next(self):
if self.prefetch:
torch.cuda.current_stream().wait_stream(self.stream)
samples = self.next_samples
targets = self.next_targets
if samples is not None:
for k, v in samples.items():
v.record_stream(torch.cuda.current_stream())
if targets is not None:
for t in targets:
for k, v in t.items():
v.record_stream(torch.cuda.current_stream())
self.preload()
else:
try:
samples, targets = next(self.loader)
samples, targets = to_cuda(samples, targets, self.device)
except StopIteration:
samples = None
targets = None
return samples, targets