def generate(self):
model_path = os.path.expanduser(self.model_path)
assert model_path.endswith(
'.h5'), 'Keras model or weights must be a .h5 file.'
self.yolo_model = yolo_body(
[self.input_shape[0], self.input_shape[1], 3], self.anchors_mask,
self.num_classes, self.phi)
self.yolo_model.load_weights(self.model_path)
print('{} model, anchors, and classes loaded.'.format(model_path))
#---------------------------------------------------------#
# 在yolo_eval函数中，我们会对预测结果进行后处理
# 后处理的内容包括，解码、非极大抑制、门限筛选等
#---------------------------------------------------------#
boxes, scores, classes = DecodeBox(
self.yolo_model.output,
self.anchors,
self.num_classes,
self.input_image_shape,
self.input_shape,
anchor_mask=self.anchors_mask,
max_boxes=self.max_boxes,
confidence=self.confidence,
nms_iou=self.nms_iou,
letterbox_image=self.letterbox_image)
return boxes, scores, classes
def train_by_fit(optimizer, loss, train_data, train_steps, validation_data,
validation_steps):
"""
使用fit方式训练，可以知道训练完的时间，以及更规范的添加callbacks参数
:param optimizer: 优化器
:param loss: 自定义的loss function
:param train_data: 以tf.data封装好的训练集数据
:param validation_data: 验证集数据
:param train_steps: 迭代一个epoch的轮次
:param validation_steps: 同上
:return: None
"""
cbk = [
callbacks.ReduceLROnPlateau(verbose=1),
callbacks.EarlyStopping(patience=10, verbose=1),
callbacks.ModelCheckpoint('./model/yolov3_{val_loss:.04f}.h5',
save_best_only=True,
save_weights_only=True)
]
model = yolo_body()
model.compile(optimizer=optimizer, loss=loss)
# initial_epoch用于恢复之前的训练
model.fit(train_data,
steps_per_epoch=max(1, train_steps),
validation_data=validation_data,
validation_steps=max(1, validation_steps),
epochs=cfg.epochs,
callbacks=cbk)
def generate(self):
model_path = os.path.expanduser(self.model_path)
assert model_path.endswith('.h5'), 'Keras model or weights must be a .h5 file.'
self.model = yolo_body([None, None, 3], self.anchors_mask, self.num_classes, self.backbone, self.alpha)
self.model.load_weights(self.model_path)
print('{} model, anchors, and classes loaded.'.format(model_path))
#---------------------------------------------------------#
# 在DecodeBox函数中，我们会对预测结果进行后处理
# 后处理的内容包括，解码、非极大抑制、门限筛选等
#---------------------------------------------------------#
self.input_image_shape = Input([2,],batch_size=1)
inputs = [*self.model.output, self.input_image_shape]
outputs = Lambda(
DecodeBox,
output_shape = (1,),
name = 'yolo_eval',
arguments = {
'anchors' : self.anchors,
'num_classes' : self.num_classes,
'input_shape' : self.input_shape,
'anchor_mask' : self.anchors_mask,
'confidence' : self.confidence,
'nms_iou' : self.nms_iou,
'max_boxes' : self.max_boxes,
'letterbox_image' : self.letterbox_image
}
)(inputs)
self.yolo_model = Model([self.model.input, self.input_image_shape], outputs)
def __init__(self, model_path):
self.class_names = cfg.class_names
self.anchors = cfg.anchors
self.score = 0.5
model = yolo_body()
model.load_weights(model_path)
self.model = model
#----------------------------------------------------#
# 获取classes和anchor
#----------------------------------------------------#
class_names, num_classes = get_classes(classes_path)
anchors, num_anchors = get_anchors(anchors_path)
#----------------------------------------------------#
# 判断是否多GPU载入模型和预训练权重
#----------------------------------------------------#
if ngpus_per_node > 1:
with strategy.scope():
#------------------------------------------------------#
# 创建yolo模型
#------------------------------------------------------#
model_body = yolo_body((None, None, 3), anchors_mask, num_classes, backbone, alpha, weight_decay=weight_decay)
if model_path != '':
#------------------------------------------------------#
# 载入预训练权重
#------------------------------------------------------#
print('Load weights {}.'.format(model_path))
model_body.load_weights(model_path, by_name=True, skip_mismatch=True)
if not eager:
model = get_train_model(model_body, input_shape, num_classes, anchors, anchors_mask, label_smoothing, focal_loss, focal_alpha, focal_gamma)
else:
#------------------------------------------------------#
# 创建yolo模型
#------------------------------------------------------#
model_body = yolo_body((None, None, 3), anchors_mask, num_classes, backbone, alpha, weight_decay=weight_decay)
if model_path != '':
#------------------------------------------------------#
os.environ["CUDA_VISIBLE_DEVICES"] = ','.join(str(x) for x in train_gpu)
ngpus_per_node = len(train_gpu)
print('Number of devices: {}'.format(ngpus_per_node))
#----------------------------------------------------#
# 获取classes和anchor
#----------------------------------------------------#
class_names, num_classes = get_classes(classes_path)
anchors, num_anchors = get_anchors(anchors_path)
K.clear_session()
#------------------------------------------------------#
# 创建yolo模型
#------------------------------------------------------#
model_body = yolo_body((input_shape[0], input_shape[1], 3),
anchors_mask,
num_classes,
phi=phi)
if model_path != '':
#------------------------------------------------------#
# 载入预训练权重
#------------------------------------------------------#
print('Load weights {}.'.format(model_path))
model_body.load_weights(model_path, by_name=True, skip_mismatch=True)
if ngpus_per_node > 1:
model = multi_gpu_model(model_body, gpus=ngpus_per_node)
model = get_train_model(model, input_shape, num_classes, anchors,
anchors_mask, label_smoothing)
else:
model = get_train_model(model_body, input_shape, num_classes, anchors,
anchors_mask, label_smoothing)
#------------------------------------------------------#
os.environ["CUDA_VISIBLE_DEVICES"] = ','.join(str(x) for x in train_gpu)
ngpus_per_node = len(train_gpu)
print('Number of devices: {}'.format(ngpus_per_node))
#----------------------------------------------------#
# 获取classes和anchor
#----------------------------------------------------#
class_names, num_classes = get_classes(classes_path)
anchors, num_anchors = get_anchors(anchors_path)
K.clear_session()
#------------------------------------------------------#
# 创建yolo模型
#------------------------------------------------------#
model_body = yolo_body((None, None, 3), anchors_mask, num_classes)
if model_path != '':
#------------------------------------------------------#
# 载入预训练权重
#------------------------------------------------------#
print('Load weights {}.'.format(model_path))
model_body.load_weights(model_path, by_name=True, skip_mismatch=True)
if ngpus_per_node > 1:
model = multi_gpu_model(model_body, gpus=ngpus_per_node)
model = get_train_model(model, input_shape, num_classes, anchors,
anchors_mask, label_smoothing, focal_loss,
focal_alpha, focal_gamma)
else:
model = get_train_model(model_body, input_shape, num_classes, anchors,
anchors_mask, label_smoothing, focal_loss,
#--------------------------------------------#
# 该部分代码用于看网络结构
#--------------------------------------------#
from nets.yolo import yolo_body
if __name__ == "__main__":
input_shape = [416, 416, 3]
anchors_mask = [[6, 7, 8], [3, 4, 5], [0, 1, 2]]
num_classes = 80
model = yolo_body(input_shape, anchors_mask, num_classes)
model.summary()
for i, layer in enumerate(model.layers):
print(i, layer.name)
#--------------------------------------------#
# 该部分代码用于看网络结构
#--------------------------------------------#
from nets.yolo import yolo_body
from utils.utils import net_flops
if __name__ == "__main__":
input_shape = [416, 416, 3]
anchors_mask = [[3, 4, 5], [1, 2, 3]]
num_classes = 80
phi = 0
model = yolo_body([input_shape[0], input_shape[1], 3], anchors_mask, num_classes, phi=phi)
#--------------------------------------------#
# 查看网络结构网络结构
#--------------------------------------------#
model.summary()
#--------------------------------------------#
# 计算网络的FLOPS
#--------------------------------------------#
net_flops(model, table=False)
#--------------------------------------------#
# 获得网络每个层的名称与序号
#--------------------------------------------#
# for i,layer in enumerate(model.layers):
# print(i,layer.name)
#--------------------------------------------#
# 该部分代码用于看网络结构
#--------------------------------------------#
from nets.yolo import yolo_body
from utils.utils import net_flops
if __name__ == "__main__":
input_shape = [416, 416]
anchors_mask = [[6, 7, 8], [3, 4, 5], [0, 1, 2]]
num_classes = 80
backbone = 'mobilenetv1'
model = yolo_body([input_shape[0], input_shape[1], 3],
anchors_mask,
num_classes,
backbone=backbone)
#--------------------------------------------#
# 查看网络结构网络结构
#--------------------------------------------#
model.summary()
#--------------------------------------------#
# 计算网络的FLOPS
#--------------------------------------------#
net_flops(model, table=False)
#--------------------------------------------#
# 获得网络每个层的名称与序号
#--------------------------------------------#
# for i,layer in enumerate(model.layers):
# print(i,layer.name)
