注意OCR - 在C ++上实现

Question

我训练了我的模型和冻结图形使用工具：tensorflow的freeze_graph，我尝试用这段代码在窗口上实现：

Session* sess = (Session*)_session;
const tensorflow::Tensor t_input = get_tensor(img_data, wImg, hImg, byte_per_line);`

    std::vector<tensorflow::Tensor> outputs;

    std::vector<std::pair<string, Tensor>> feed_dict = {
        { "input_layers:0", t_input },
    };

    const std::string nodeOutPut = "AttentionOcr_v1/predicted_chars";
    Status run_status = sess->Run(feed_dict, { nodeOutPut },
        {}, &outputs);
    if (!run_status.ok()) {
        return;
    }

t_input形状为：[1, w, h, 3] 但它返回：

run_status = {state_=0x0000000004958cb0 {code=INVALID_ARGUMENT (3) msg=<Error reading characters of string.> } }

，任何人都可以帮助我，提前谢谢