frugally-deep: Header-only library for using Keras models in C++

    // Convenience wrapper around predict for models with

    // single tensor outputs of shape (1, 1, 1),

    // typically used for regression or binary classification.

    // Returns this one activation value.

    float_type predict_single_output(const tensor5s& inputs) const

    {

        const tensor5s outputs = predict(inputs);

        internal::assertion(outputs.size() == 1,

            "invalid number of outputs");

        const auto output_shape = outputs.front().shape();

        internal::assertion(output_shape.volume() == 1,

            "invalid output shape");

        return outputs.front().get(0, 0, 0, 0, 0);

}

convert tensor5 datatype to std::vector

const std::vector<float> result_vec = *result.front().as_vector();

巴特西

frugally-deep: Header-only library for using Keras models in C++

最新文章

热门文章