google_deeplab代码分析

Posted on

vis.py : 

1
2
predictions = model.predict_labels(
)

->model.py

def predict_labels()

1
2
outputs_to_scales_to_logits = multi_scale_logits(
)

def multi_scale_logits()

1
outputs_to_logits = _get_logits()

def _get_logits()

1
features, end_point = extract_features()

def extract_features()

->core/feature_extractor.py

def extract_features()

model_variant [‘resnet’, ‘xception’, ‘mobilenet’, ‘nas’]

1
2
arg_scope = arg_scopes_map[model_variant]()
features, end_point = get_network(model_variant,..., arg_scope)

model_variant=’xception-XX’

-> arg_scopes_map[model_variant]=xception_arg_scope

->core/xception.py

def xception_arg_scope() 不知道干啥的

def get_network()

1
func = network_map[network_name]

->core/xception.py

def xception_71()