def combined_roidb(imdb_names):
"""
Combine multiple roidbs
"""
def get_roidb(imdb_name):
# 构造了两个类：数据集类pascal_voc和图片类imdb
imdb = get_imdb(imdb_name)
print('Loaded dataset `{:s}` for training'.format(imdb.name))
# 即选择真实物体groundtruth
imdb.set_proposal_method("gt")
print('Set proposal method: {:s}'.format("gt"))
# 获得训练数据
roidb = get_training_roidb(imdb)
return roidb
roidbs = [get_roidb(s) for s in imdb_names.split('+')]
roidb = roidbs[0]
if len(roidbs) > 1:
for r in roidbs[1:]:
roidb.extend(r)
tmp = get_imdb(imdb_names.split('+')[1])
imdb = imdb2(imdb_names, tmp.classes)
else:
imdb = get_imdb(imdb_names)
return imdb, roidb
def combined_roidb(imdb_names):
"""
Combine multiple roidbs
"""
#print('get_roidb')
def get_roidb(imdb_name):
imdb = get_imdb(imdb_name)
print('Loaded Dataset `{:s}` for training'.format(imdb.name))
imdb.set_proposal_method("gt")
print('Set proposal method: {:s}'.format("gt"))
roidb = get_training_roidb(imdb)
return roidb
roidbs = [get_roidb(s) for s in imdb_names.split('+')]
roidb = roidbs[0]
#print('roidbs',roidbs)
if len(roidbs) > 1:
for r in roidbs[1:]:
roidb.extend(r)
tmp = get_imdb(imdb_names.split('+')[1])
imdb = imdb2(imdb_names, tmp.classes)
else:
imdb = get_imdb(imdb_names)
return imdb, roidb
def combined_roidb(imdb_names):
"""Combine multiple roidbs"""
# 融合roidb，roidb来自于数据集（实验可能用到多个），所以需要combine多个数据集的roidb
def get_roidb(imdb_name):
imdb = get_imdb(imdb_name)
print('imdb*******/n', imdb)
print('Loaded dataset `{:s}` for training'.format(imdb_name))
imdb.set_proposal_method("gt")
# 设置proposal方法
print('Set proposal method: gt')
roidb = get_training_roidb(imdb)
# 得到用于训练的roidb，定义在train.py，进行了水平翻转，以及为原始roidb添加了一些说明性的属性
return roidb
roidbs = [get_roidb(s) for s in imdb_names.split('+')]
roidb = roidbs[0]
if len(roidbs) > 1:
# 如果大于一个，则进行combine roidb
for r in roidbs[1:]:
roidb.extend(r)
tmp = get_imdb(imdb_names.split('+')[1])
imdb = imdb2(imdb_names, tmp.classes)
else:
imdb = get_imdb(imdb_names)
# get_imdb方法定义在factory.py中，通过名字得到imdb
return imdb, roidb
def combined_roidb(imdb_names):
"""
Combine multiple roidbs
"""
def get_roidb(imdb_name):
#实例化一个pascal_voc类对象
imdb = get_imdb(imdb_name)
print('Loaded dataset `{:s}` for training'.format(imdb.name))
#为imbd.roidb_handler赋值为self.gt_roidb
imdb.set_proposal_method("gt")
print('Set proposal method: {:s}'.format("gt"))
roidb = get_training_roidb(imdb)
return roidb
roidbs = [get_roidb(s) for s in imdb_names.split('+')]
roidb = roidbs[0]
if len(roidbs) > 1:
for r in roidbs[1:]:
roidb.extend(r)
tmp = get_imdb(imdb_names.split('+')[1])
imdb = imdb2(imdb_names, tmp.classes)
else:
#实例化一个pascal_voc类对象
imdb = get_imdb(imdb_names)
return imdb, roidb
def combined_roidb(imdb_names):
"""
Combine multiple roidbs
"""
if imdb_names.split('+').__len__() > 1:
tmp = get_imdb(imdb_names.split('+')[1])
imdb = imdb2(imdb_names, tmp.classes)
else:
imdb = get_imdb(imdb_names)
return imdb
def combined_roidb(imdb_names):
"""
Combine multiple roidbs
"""
roidbs = [get_roidb(s) for s in imdb_names.split('+')]
roidb = roidbs[0]
if len(roidbs) > 1:
for r in roidbs[1:]:
roidb.extend(r)
tmp = get_imdb(imdb_names.split('+')[1])
imdb = imdb2(imdb_names, tmp.classes)
else:
imdb = get_imdb(imdb_names)
return imdb, roidb
def combined_roidb(imdb_names, training=True, root_path=None):
"""
Combine multiple roidbs
"""
def get_training_roidb(imdb):
"""Returns a roidb (Region of Interest database) for use in training."""
if cfg.TRAIN.USE_FLIPPED:
print('Appending horizontally-flipped training examples...')
imdb.append_flipped_images()
if cfg.TRAIN.USE_VER_FLIPPED:
print('Appending vertically-flipped training examples...')
imdb.append_vertical_flipped_images()
print('Preparing training data...')
prepare_roidb(imdb)
# ratio_index = rank_roidb_ratio(imdb)
return imdb.roidb
def get_roidb(imdb_name, root_path=None):
imdb = get_imdb(imdb_name, root_path)
print('Loaded dataset `{:s}` for training'.format(imdb.name))
imdb.set_proposal_method(cfg.TRAIN.PROPOSAL_METHOD)
print('Set proposal method: {:s}'.format(cfg.TRAIN.PROPOSAL_METHOD))
roidb = get_training_roidb(imdb)
return roidb
roidbs = [get_roidb(s, root_path=root_path) for s in imdb_names.split('+')]
roidb = roidbs[0]
if len(roidbs) > 1:
for r in roidbs[1:]:
roidb.extend(r)
tmp = get_imdb(imdb_names.split('+')[1], root_path)
imdb = datasets.imdb.imdb(imdb_names, tmp.classes)
else:
imdb = get_imdb(imdb_names, root_path)
if training:
roidb = filter_roidb(roidb)
return imdb, roidb
else:
ratio_list, ratio_index = rank_roidb_ratio(roidb)
return imdb, roidb, ratio_list, ratio_index
def get_roidb(imdb_name):
imdb = get_imdb(imdb_name)
print('Loaded dataset `{:s}` for training'.format(imdb.name))
imdb.set_proposal_method(cfg.TRAIN.PROPOSAL_METHOD)
print('Set proposal method: {:s}'.format(cfg.TRAIN.PROPOSAL_METHOD))
roidb = get_training_roidb(imdb)
return roidb
def load_db(db_name):
"""
TOR-related database
"""
roidb = get_roidb(db_name)
imdb = get_imdb(db_name)
return imdb, roidb
def train_main(data_dir, model_dir, train_steps, input_yaml):
cfg_from_file(input_yaml)
print('Using config:')
pprint.pprint(cfg)
imdb = get_imdb('voc_2007_trainval')
print('Loaded dataset `{:s}` for training'.format(imdb.name))
roidb = get_training_roidb(imdb)
output_dir = get_output_dir(imdb, None)
log_dir = get_log_dir(imdb)
print('Output will be saved to `{:s}`'.format(output_dir))
print('Logs will be saved to `{:s}`'.format(log_dir))
device_name = '/gpu:0'
print(device_name)
network = get_network('VGGnet_train')
train_net(network,
imdb,
roidb,
output_dir=output_dir,
log_dir=log_dir,
pretrained_model='data/pretrain_model/VGG_imagenet.npy',
max_iters=int(cfg.TRAIN.max_steps),
restore=bool(int(cfg.TRAIN.restore)))
def get_roidb(imdb_name):
imdb = get_imdb(imdb_name)
print('Loaded dataset `{:s}` for training'.format(imdb.name))
imdb.set_proposal_method("gt")
print('Set proposal method: {:s}'.format("gt"))
roidb = get_training_roidb(imdb)
return roidb
def get_roidb(imdb_name):
#imdb_name=voc_2007_trainval
imdb = get_imdb(imdb_name)#imdb = pascal_voc(trainval, 2007) pascal_voc是imdb的子类
print('Loaded dataset `{:s}` for training'.format(imdb.name))#imdb.name= voc_2007_trainval
imdb.set_proposal_method(cfg.TRAIN.PROPOSAL_METHOD) #TRAIN.PROPOSAL_METHOD='gt'
print('Set proposal method: {:s}'.format(cfg.TRAIN.PROPOSAL_METHOD))#'gt'
roidb = get_training_roidb(imdb)
return roidb
def get_roidb(imdb_name):
# 调用factory文件中的get_imdb()方法
imdb = get_imdb(imdb_name) # imdb为调用get_imdb返回的pascal voc类对象
print('Loaded dataset {:s}'.format(imdb.name))
imdb.set_proposal_method(cfg.TRAIN.PROPOSAL_METHOD) # gt
print('Set proposal method: {:s}'.format(cfg.TRAIN.PROPOSAL_METHOD))
roidb = get_training_roidb(imdb)
return roidb
def get_roidb(imdb_name):
# 初始化了imdb，已经获取了一些基础配置
imdb = get_imdb(imdb_name)
print('Loaded dataset `{:s}` for training'.format(imdb.name))
imdb.set_proposal_method("gt")
print('Set proposal method: {:s}'.format("gt"))
# 为imdb.roidb添加了一些配置，宽高，最大类别等
roidb = get_training_roidb(imdb)
return roidb
def get_roidb(imdb_name): #本函数根据数据库名字 获取roidb
imdb = get_imdb(imdb_name) #调用factory 的 imdb 然后根据 数据库 名字 调用 pascal_voc
print('Loaded dataset `{:s}` for training'.format(imdb.name))
imdb.set_proposal_method(cfg.TRAIN.PROPOSAL_METHOD) #设置gt 的方法
print('Set proposal method: {:s}'.format(cfg.TRAIN.PROPOSAL_METHOD))
roidb = get_training_roidb(imdb)
return imdb, roidb
def get_roidb(imdb_name):
# 获取voc 原始数据集
imdb = get_imdb(imdb_name)
print('Loaded dataset `{:s}` for training'.format(imdb.name))
imdb.set_proposal_method(cfg.TRAIN.PROPOSAL_METHOD)
# print('Set proposal method: {:s}'.format(cfg.TRAIN.PROPOSAL_METHOD))
# 制作并获取目标区域的位置信息
roidb = get_training_roidb(imdb)
return roidb
def GetRoidb(imdb_name):
"""
Note: we need to run get_training_roidb sort of funcs later
for now, it only supports single roidb.
"""
imdb = get_imdb(imdb_name)
roidb = imdb.roidb
return imdb, roidb
def get_roidb(imdb_name):
#实例化一个pascal_voc类对象
imdb = get_imdb(imdb_name)
print('Loaded dataset `{:s}` for training'.format(imdb.name))
#为imbd.roidb_handler赋值为self.gt_roidb
imdb.set_proposal_method("gt")
print('Set proposal method: {:s}'.format("gt"))
roidb = get_training_roidb(imdb)
return roidb
def get_roidb(imdb_name):
# 构造了两个类：数据集类pascal_voc和图片类imdb
imdb = get_imdb(imdb_name)
print('Loaded dataset `{:s}` for training'.format(imdb.name))
# 即选择真实物体groundtruth
imdb.set_proposal_method("gt")
print('Set proposal method: {:s}'.format("gt"))
# 获得训练数据
roidb = get_training_roidb(imdb)
return roidb
def combined_roidb(imdb_names, training=True):
def get_training_roidb(imdb):
if cfg.TRAIN.USE_FLIPPED: # USE_FLIPPED = True
print('Appending horizontally-flipped training examples...')
imdb.append_flipped_images() # 水平翻转gt_boxes
print('done')
print('Prapareing training data...')
prepare_roidb(imdb) # roidb.prepare_roidb()
# prepare_roidb对每张图片的roidb进行信息扩充，添加id，路径，宽高，box类别信息等
#ratio_index = rank_roidb_ratio(imdb)
print('done')
return imdb.roidb
def get_roidb(imdb_name):
# 调用factory文件中的get_imdb()方法
imdb = get_imdb(imdb_name) # imdb为调用get_imdb返回的pascal voc类对象
print('Loaded dataset {:s}'.format(imdb.name))
imdb.set_proposal_method(cfg.TRAIN.PROPOSAL_METHOD) # gt
print('Set proposal method: {:s}'.format(cfg.TRAIN.PROPOSAL_METHOD))
roidb = get_training_roidb(imdb)
return roidb
# s = 'voc_2007_trainval' 调用get_roidb()
roidbs = [get_roidb(s) for s in imdb_names.split('+')]
roidb = roidbs[0]
if len(roidbs) > 1:
for r in roidbs[1:]:
roidb.extend(r)
tmp = get_imdb(imdb_names.split('+')[1])
imdb = datasets.imdb.imdb(imdb_names, tmp.classes)
else:
# 由于前面定义的imdb对象在get_roidb函数中，若返回该对象需再次定义
imdb = get_imdb(imdb_names)
if training:
roidb = filter_roidb(roidb) # 过滤不含box的roidb
ratio_list, ratio_index = rank_roidb_ratio(roidb)
# 返回从小到大排序的ratio_list以及对应的原图roidb像索引
return imdb, roidb, ratio_list, ratio_index
def combined_roidb(imdb_names):
def get_roidb(imdb_name):
# 初始化了imdb，已经获取了一些基础配置
imdb = get_imdb(imdb_name)
print('Loaded dataset `{:s}` for training'.format(imdb.name))
imdb.set_proposal_method("gt")
print('Set proposal method: {:s}'.format("gt"))
# 为imdb.roidb添加了一些配置，宽高，最大类别等
roidb = get_training_roidb(imdb)
return roidb
roidbs = [get_roidb(s) for s in imdb_names.split('+')]
roidb = roidbs[0]
if len(roidbs) > 1:
for r in roidbs[1:]:
roidb.extend(r)
tmp = get_imdb(imdb_names.split('+')[1])
imdb = imdb2(imdb_names, tmp.classes)
else:
imdb = get_imdb(imdb_names)
return imdb, roidb
def get_roidb(imdb_name):
imdb = get_imdb(imdb_name)
print('imdb*******/n', imdb)
print('Loaded dataset `{:s}` for training'.format(imdb_name))
imdb.set_proposal_method("gt")
# 设置proposal方法
print('Set proposal method: gt')
roidb = get_training_roidb(imdb)
# 得到用于训练的roidb，定义在train.py，进行了水平翻转，以及为原始roidb添加了一些说明性的属性
return roidb
def combined_roidb(imdb_names):
"""
Combine multiple roidbs
"""
def get_roidb(imdb_name):
imdb = get_imdb(imdb_name)
print('Loaded dataset `{:s}` for training'.format(imdb.name))
imdb.set_proposal_method(cfg.TRAIN.PROPOSAL_METHOD)
print('Set proposal method: {:s}'.format(cfg.TRAIN.PROPOSAL_METHOD))
roidb = get_training_roidb(imdb)
return roidb
roidbs = [get_roidb(s) for s in imdb_names.split('+')]
roidb = roidbs[0]
if len(roidbs) > 1:
for r in roidbs[1:]:
roidb.extend(r)
tmp = get_imdb(imdb_names.split('+')[1])
imdb = lib.datasets.imdb.imdb(imdb_names, tmp.classes)
else:
imdb = get_imdb(imdb_names)
return imdb, roidb
def get_roidb_limit_ram(imdb_name):
"""
Note: we need to run get_training_roidb sort of funcs later
for now, it only supports single roidb.
"""
imdb = get_imdb(imdb_name)
roidb = imdb.roidb
assert isinstance(roidb, six.string_types), \
"for limit ram vision, roidb should be a path."
return imdb, roidb
def load_db(db_name):
"""
TOR-related database
"""
def get_roidb(imdb_name):
imdb = get_imdb(imdb_name)
print('Loaded dataset `{:s}` for training'.format(imdb.name))
imdb.set_proposal_method("gt")
print('Set proposal method: {:s}'.format("gt"))
roidb = get_training_roidb(imdb)
return roidb
roidb = get_roidb(db_name)
imdb = get_imdb(db_name)
return imdb, roidb
def from_dets(imdb_name, output_dir, args):
imdb = get_imdb(imdb_name)
imdb.competition_mode(args.comp_mode)
imdb.config['matlab_eval'] = args.matlab_eval
with open(os.path.join(output_dir, 'detections.pkl'), 'rb') as f:
dets = pickle.load(f)
if args.apply_nms:
print('Applying NMS to all detections')
nms_dets = apply_nms(dets, cfg.TEST.NMS)
else:
nms_dets = dets
print('Evaluating detections')
imdb.evaluate_detections(nms_dets, output_dir)
def __init__(self, imdb_name, resize=True):
imdb = get_imdb(imdb_name)
# Ignore the background class!!! So ['gt_classes'] must minus 1.
self.classes = [
np.zeros(imdb.num_classes - 1) for i in range(imdb.num_images)
]
for i, anno in enumerate(imdb.gt_roidb()):
np.put(self.classes[i], map(lambda x: x - 1, anno['gt_classes']),
1)
# np.put(self.classes[i], random.choice(map(lambda x: x-1, anno['gt_classes'])), 1)
self.images = [imdb.image_path_at(i) for i in range(imdb.num_images)]
assert len(self.classes) == len(self.images)
self._perm = np.random.permutation(np.arange(len(self.images)))
self._cur = 0
self.resize = resize
def combined_roidb(imdb_names):
def get_roidb(imdb_name):
imdb = get_imdb(imdb_name)
print 'Loaded dataset `{:s}` for training'.format(imdb.name)
imdb.set_proposal_method(cfg.TRAIN.PROPOSAL_METHOD)
print 'Set proposal method: {:s}'.format(cfg.TRAIN.PROPOSAL_METHOD)
roidb = get_training_roidb(imdb)
return roidb
roidbs = [get_roidb(s) for s in imdb_names.split('+')]
roidb = roidbs[0]
if len(roidbs) > 1:
for r in roidbs[1:]:
roidb.extend(r)
imdb = datasets.imdb.imdb(imdb_names)
else:
imdb = get_imdb(imdb_names)
return imdb, roidb
print('Using config:')
pprint.pprint(cfg)
# if has model, get the name from it
# if does not, then just use the initialization weights
if args.model:
filename = os.path.splitext(os.path.basename(args.model))[0]
else:
filename = os.path.splitext(os.path.basename(args.weight))[0]
tag = args.tag
tag = tag if tag else 'default'
filename = tag + '/' + filename
imdb = get_imdb(args.imdb_name)
imdb.competition_mode(args.comp_mode)
tfconfig = tf.ConfigProto(allow_soft_placement=True)
tfconfig.gpu_options.allow_growth = True
# init session
sess = tf.Session(config=tfconfig)
# load network
if args.backbone == 'vgg16':
net = vgg16_ssh()
elif args.backbone == 'res50':
net = resnetv1_ssh(num_layers=50)
elif args.backbone == 'res101':
net = resnetv1_ssh(num_layers=101)
elif args.backbone == 'res152':
if last_snapshot_iter != self.solver.iter:
self.snapshot()
def get_training_roidb(imdb):
"""Returns a roidb (Region of Interest database) for use in training."""
if cfg.TRAIN.USE_FLIPPED:
print 'Appending horizontally-flipped training examples...'
imdb.append_flipped_images()
print 'done'
print 'Preparing training data...'
rdl_roidb.prepare_roidb(imdb)
print 'done'
return imdb.roidb
def train_net(solver_prototxt, roidb, output_dir,
pretrained_model=None, max_iters=40000):
"""Train a Fast R-CNN network."""
sw = SolverWrapper(solver_prototxt, roidb, output_dir,
pretrained_model=pretrained_model)
print 'Solving...'
sw.train_model(max_iters)
print 'done solving'
if __name__ == '__main__':
from lib.datasets.factory import get_imdb
a = train_net('models/VGG_CNN_M_1024/solver.prototxt', get_training_roidb(get_imdb('soton_ear')), output_dir=None)
pass
import os.path
sys.path.append(os.getcwd())
this_dir = os.path.dirname(__file__)
from lib.fast_rcnn.train import get_training_roidb, train_net
from lib.fast_rcnn.config import cfg_from_file, get_output_dir, get_log_dir
from lib.datasets.factory import get_imdb
from lib.networks.factory import get_network
from lib.fast_rcnn.config import cfg
if __name__ == '__main__':
cfg_from_file('ctpn/text.yml')
print('Using config:')
pprint.pprint(cfg)
imdb = get_imdb('voc_2007_trainval')
print('Loaded dataset `{:s}` for training'.format(imdb.name))
roidb = get_training_roidb(imdb)
output_dir = get_output_dir(imdb, None)
log_dir = get_log_dir(imdb)
print('Output will be saved to `{:s}`'.format(output_dir))
print('Logs will be saved to `{:s}`'.format(log_dir))
device_name = '/gpu:0'
print(device_name)
network = get_network('VGGnet_train')
train_net(network,
imdb,
print('Called with args:')
print(args)
if args.cfg_file is not None:
cfg_from_file(args.cfg_file)
if args.set_cfgs is not None:
cfg_from_list(args.set_cfgs)
print('Using config:')
pprint.pprint(cfg)
if not args.randomize:
# fix the random seeds (numpy and caffe) for reproducibility
np.random.seed(cfg.RNG_SEED)
imdb = get_imdb(args.imdb_name)
print 'Loaded dataset `{:s}` for training'.format(imdb.name)
roidb = get_training_roidb(imdb)
output_dir = get_output_dir(imdb, None)
log_dir = get_log_dir(imdb)
print 'Output will be saved to `{:s}`'.format(output_dir)
print 'Logs will be saved to `{:s}`'.format(log_dir)
device_name = '/gpu:{:d}'.format(args.gpu_id)
print device_name
network = get_network(args.network_name)
print 'Use network `{:s}` in training'.format(args.network_name)
train_net(network, imdb, roidb,
import os.path
sys.path.append(os.getcwd())
this_dir = os.path.dirname(__file__)
from lib.fast_rcnn.train import get_training_roidb, train_net
from lib.fast_rcnn.config import cfg_from_file, get_output_dir, get_log_dir
from lib.datasets.factory import get_imdb
from lib.networks.factory import get_network
from lib.fast_rcnn.config import cfg
if __name__ == '__main__':
cfg_from_file('ctpn/text.yml')
print('Using config:')
pprint.pprint(cfg)
imdb = get_imdb('voc_2007_trainval')
print('Loaded dataset `{:s}` for training'.format(imdb.name))
roidb = get_training_roidb(imdb)
output_dir = get_output_dir(imdb, None)
log_dir = get_log_dir(imdb)
print('Output will be saved to `{:s}`'.format(output_dir))
print('Logs will be saved to `{:s}`'.format(log_dir))
device_name = '/gpu:0'
print(device_name)
network = get_network('VGGnet_train')
train_net(network, imdb, roidb,
output_dir=output_dir,
