def save_cate(cate_list, image_dir, res_dir):
"""
保存裁剪的图像
"""
# 保存image_dir下所有图像，以便后续查询
image_dict = dict()
image_name_list = os.listdir(image_dir)
for name in image_name_list:
image_path = os.path.join(image_dir, name)
img = cv2.imread(image_path)
image_dict[name.split('.')[0]] = img
# 遍历所有类别，保存标注的图像
for i in range(20):
cate_name = alphabets[i]
cate_dir = os.path.join(res_dir, cate_name)
util.check_dir(cate_dir)
for item in cate_list[i]:
img_name = item['img_name']
xmin, ymin, xmax, ymax = item['rect']
rect_img = image_dict[img_name][ymin:ymax, xmin:xmax]
img_path = os.path.join(
cate_dir,
'%s-%d-%d-%d-%d.png' % (img_name, xmin, ymin, xmax, ymax))
cv2.imwrite(img_path, rect_img)
def _resume_checkpoint(self, path):
check_dir(path)
self.logger.debug('Loading checkpoint: {}...'.format(path))
checkpoint = torch.load(path)
self.start_epoch = checkpoint['epoch'] + 1
self.best_score = checkpoint['best_score']
self.model.load_state_dict(checkpoint['state_dict'])
if checkpoint['config']['optimizer'] != self.cfg['optimizer']:
self.logger.debug("Optimizer type given in config file is different from that of checkpoint."
"Optimizer parameters not being resumed.")
else:
self.optimizer.load_state_dict(checkpoint['optimizer'])
self.logger.debug("Checkpoint loaded. Resume training from epoch {}".format(self.start_epoch))
def record_voice(files, base_path):
for file_path in files:
play_audio(file_path)
audio = pyaudio.PyAudio()
stream = audio.open(format=pyaudio.paInt16,
channels=1,
rate=16000,
input=True,
frames_per_buffer=SIZE)
names = file_path.split('/')
print("--- ", end="")
print(os.path.join(base_path, names[1], names[2]))
frames = []
for i in range(
0, int(16000 / SIZE * get_duration('dataset/' + file_path))):
data = stream.read(SIZE)
frames.append(data)
stream.stop_stream()
stream.close()
audio.terminate()
util.check_dir(path=base_path + names[1])
util.save_wav(frames,
file_path=os.path.join(base_path, names[1], names[2]))
criterion = SmoothLabelCritierion(label_smoothing=0.1)
# criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=1e-3, weight_decay=1e-4)
scheduler = optim.lr_scheduler.CosineAnnealingLR(optimizer,
num_epochs - 5,
eta_min=1e-4)
lr_scheduler = GradualWarmupScheduler(optimizer,
multiplier=1,
total_epoch=5,
after_scheduler=scheduler)
optimizer.zero_grad()
optimizer.step()
lr_scheduler.step()
util.check_dir('../data/models/')
best_model, loss_dict, top1_acc_dict, top5_acc_dict = train_model(
data_loaders,
data_sizes,
name,
model,
criterion,
optimizer,
lr_scheduler,
num_epochs=num_epochs,
device=device)
# 保存最好的模型参数
# util.save_model(best_model.cpu(), '../data/models/best_%s.pth' % name)
res_loss[name] = loss_dict
res_top1_acc[name] = top1_acc_dict
if 0.1 <= iou_score < 0.5:
# 负样本
negative_list.append(rects[i])
else:
pass
# 添加标注边界框到正样本列表
positive_list.extend(bndboxs)
return positive_list, negative_list
if __name__ == '__main__':
car_root_dir = '../../data/voc_car/'
finetune_root_dir = '../../data/finetune_car/'
check_dir(finetune_root_dir)
gs = selectivesearch.get_selective_search()
for name in ['train', 'val']:
src_root_dir = os.path.join(car_root_dir, name)
src_annotation_dir = os.path.join(src_root_dir, 'Annotations')
src_jpeg_dir = os.path.join(src_root_dir, 'JPEGImages')
dst_root_dir = os.path.join(finetune_root_dir, name)
dst_annotation_dir = os.path.join(dst_root_dir, 'Annotations')
dst_jpeg_dir = os.path.join(dst_root_dir, 'JPEGImages')
check_dir(dst_root_dir)
check_dir(dst_annotation_dir)
check_dir(dst_jpeg_dir)
total_num_positive = 0
model = model.to(device)
criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=1e-3)
lr_scheduler = optim.lr_scheduler.StepLR(optimizer,
step_size=15,
gamma=0.1)
best_model, loss_dict, acc_dict = train_model(model,
criterion,
optimizer,
lr_scheduler,
data_sizes,
data_loaders,
num_epochs=50,
device=device)
# 保存最好的模型参数
util.check_dir(model_dir)
torch.save(best_model.state_dict(),
os.path.join(model_dir, '%s.pth' % name))
res_loss[name] = loss_dict
res_acc[name] = acc_dict
print('train %s done' % name)
print()
util.save_png('loss', res_loss)
util.save_png('acc', res_acc)
src_jpeg_path = os.path.join(voc_jpeg_dir, sample_name + suffix_jpeg)
dst_jpeg_path = os.path.join(data_jpeg_dir, sample_name + suffix_jpeg)
shutil.copyfile(src_jpeg_path, dst_jpeg_path)
csv_path = os.path.join(data_root_dir, 'car.csv')
np.savetxt(csv_path, np.array(car_samples), fmt='%s')
if __name__ == '__main__':
samples = {
'train': parse_train_val(car_train_path),
'val': parse_train_val(car_val_path)
}
print(samples)
# samples = sample_train_val(samples)
# print(samples)
check_dir(car_root_dir)
for name in ['train', 'val']:
data_root_dir = os.path.join(car_root_dir, name)
data_annotation_dir = os.path.join(data_root_dir, 'Annotations')
data_jpeg_dir = os.path.join(data_root_dir, 'JPEGImages')
check_dir(data_root_dir)
check_dir(data_annotation_dir)
check_dir(data_jpeg_dir)
save_car(samples[name], data_root_dir, data_annotation_dir,
data_jpeg_dir)
print('done')
cate_dir = os.path.join(res_dir, cate_name)
util.check_dir(cate_dir)
for item in cate_list[i]:
img_name = item['img_name']
xmin, ymin, xmax, ymax = item['rect']
rect_img = image_dict[img_name][ymin:ymax, xmin:xmax]
img_path = os.path.join(
cate_dir,
'%s-%d-%d-%d-%d.png' % (img_name, xmin, ymin, xmax, ymax))
cv2.imwrite(img_path, rect_img)
if __name__ == '__main__':
util.check_dir(root_dir)
util.check_dir(train_dir)
util.check_dir(val_dir)
train_name_list = np.loadtxt(train_txt_path, dtype=np.str)
print(train_name_list)
cate_list = find_all_cate_rects(annotation_dir, train_name_list)
print([len(x) for x in cate_list])
save_cate(cate_list, jpeg_image_dir, train_dir)
val_name_list = np.loadtxt(val_txt_path, dtype=np.str)
print(val_name_list)
cate_list = find_all_cate_rects(annotation_dir, val_name_list)
print([len(x) for x in cate_list])
save_cate(cate_list, jpeg_image_dir, val_dir)
rect_size = (ymax - ymin) * (xmax - xmin)
iou_score = iou_list[i]
if 0 < iou_score <= 0.3 and rect_size > maximum_bndbox_size / 5.0:
# 负样本
negative_list.append(rects[i])
else:
pass
return bndboxs, negative_list
if __name__ == '__main__':
car_root_dir = '../../data/voc_car/'
classifier_root_dir = '../../data/classifier_car/'
check_dir(classifier_root_dir)
gs = selectivesearch.get_selective_search()
for name in ['train', 'val']:
src_root_dir = os.path.join(car_root_dir, name)
src_annotation_dir = os.path.join(src_root_dir, 'Annotations')
src_jpeg_dir = os.path.join(src_root_dir, 'JPEGImages')
dst_root_dir = os.path.join(classifier_root_dir, name)
dst_annotation_dir = os.path.join(dst_root_dir, 'Annotations')
dst_jpeg_dir = os.path.join(dst_root_dir, 'JPEGImages')
check_dir(dst_root_dir)
check_dir(dst_annotation_dir)
check_dir(dst_jpeg_dir)
total_num_positive = 0
model.load_state_dict(best_model_weights)
return model
if __name__ == '__main__':
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
data_loaders, data_sizes = load_data('./data/finetune_car')
model = models.alexnet(pretrained=True)
# print(model)
num_features = model.classifier[6].in_features
model.classifier[6] = nn.Linear(num_features, 2)
# print(model)
model = model.to(device)
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=1e-3, momentum=0.9)
lr_scheduler = optim.lr_scheduler.StepLR(optimizer, step_size=7, gamma=0.1)
best_model = train_model(data_loaders,
model,
criterion,
optimizer,
lr_scheduler,
device=device,
num_epochs=25)
# 保存最好的模型参数
check_dir('./models')
torch.save(best_model.state_dict(), 'models/alexnet_car.pth')
"""
voc_car_train_dir = '../../data/voc_car/train'
# ground truth
gt_annotation_dir = os.path.join(voc_car_train_dir, 'Annotations')
jpeg_dir = os.path.join(voc_car_train_dir, 'JPEGImages')
classifier_car_train_dir = '../../data/finetune_car/train'
# positive
positive_annotation_dir = os.path.join(classifier_car_train_dir, 'Annotations')
dst_root_dir = '../../data/bbox_regression/'
dst_jpeg_dir = os.path.join(dst_root_dir, 'JPEGImages')
dst_bndbox_dir = os.path.join(dst_root_dir, 'bndboxs')
dst_positive_dir = os.path.join(dst_root_dir, 'positive')
util.check_dir(dst_root_dir)
util.check_dir(dst_jpeg_dir)
util.check_dir(dst_bndbox_dir)
util.check_dir(dst_positive_dir)
samples = util.parse_car_csv(voc_car_train_dir)
res_samples = list()
total_positive_num = 0
for sample_name in samples:
# 提取正样本边界框坐标（IoU>=0.5）
positive_annotation_path = os.path.join(positive_annotation_dir, sample_name + '_1.csv')
positive_bndboxes = np.loadtxt(positive_annotation_path, dtype=np.int, delimiter=' ')
# 提取标注边界框
gt_annotation_path = os.path.join(gt_annotation_dir, sample_name + '.xml')
bndboxs = util.parse_xml(gt_annotation_path)
# 计算符合条件（IoU>0.6）的候选建议