深度學習模型在移動端部署方法

1. Keras 轉 tflite

def keras2tflite(keras_model, tflitefile)
	converter = tf.lite.TFLiteConverter.from_keras_model(model_2)
	# Indicate that we want to perform the default optimizations,
	converter.optimizations = [tf.lite.Optimize.DEFAULT]
	tflite_model = converter.convert()
	open(tflitefile, "wb").write(tflite_model)

　　模型轉化之后，就可以在移動端通過tflite相應的工具進行調用了。比如，如果需要通過c語言的調用的話，則可以將得到的模型轉化為c語言的代碼。

apt-get -qq install xxd
xxd -i model.tflite > model.cc
cat model.cc  #查看生成的模型文件

2. Pytorch to Onnx

def pytorch2onnx(model, onnxfile,cpu = True):
    device = torch.device("cpu" if cpu else "cuda")
    net = model.to(device)
    inputs = torch.randn(1, 3, WIDTH, HEIGHT).to(device)
    torch_out = torch.onnx._export(net, inputs, output_onnx, export_params=True, verbose=False)