TensorFlow h5文件转onnx

前期工作，把各个包安装好

1. h5转pb
   这里采用的是先读取模型，然后再将模型转为pb格式

def h5_to_pb(model):
    # model = retinaface_model.build_model()

    model.summary()

    full_model = tf.function(lambda Input: model(Input))

    full_model = full_model.get_concrete_function(tf.TensorSpec(model.inputs[0].shape, model.inputs[0].dtype))

    # Get frozen ConcreteFunction

    frozen_func = convert_variables_to_constants_v2(full_model)

    frozen_func.graph.as_graph_def()

    layers = [op.name for op in frozen_func.graph.get_operations()]

    print("-" * 50)

    print("Frozen model layers: ")

    for layer in layers:
        print(layer)

    tf.io.write_graph(graph_or_graph_def=frozen_func.graph,
                      logdir="./models",
                      name="model.pb",
                      as_text=False)

2. pb转onnx
   如果只有一个输出可以不需要那么多output，只需要附上最后的节点即可
   可以通过以下函数查看输入输出节点名称

def create_graph(path):
    with tf.compat.v1.gfile.FastGFile(path, 'rb') as f:
        graph_def = tf.compat.v1.GraphDef()
        graph_def.ParseFromString(f.read())
        tf.import_graph_def(graph_def, name='')
    tensor_name_list = [tensor.name for tensor in tf.compat.v1.get_default_graph().as_graph_def().node]

    print([tensor.name for tensor in tf.compat.v1.get_default_graph().as_graph_def().node])
    # print("input_name:" + tensor_name_list[0])
    # print("output_name:" + tensor_name_list[-1])

python -m tf2onnx.convert --input ./models/model.pb --inputs Input:0 --output ./models/model.onnx --outputs Identity:0,Identity_1:0,Identity_2:0,Identity_3:0,Ident
ity_4:0,Identity_5:0,Identity_6:0,Identity_7:0,Identity_8:0 --opset 11