How to speed up the python side's access to C++ side std::vector<T>

[LeBron-Jian]

Firstly, I create a function Dataset which takes a std::vector as input, and i would like to expose the function to python， The function is constructed as follows

struct Label{
    int x;
    int y;
    double w;
    double h;
}

class Dataset {
public:
    std::vector<std::wstring>            imgs;
    std::vector<int>                     ids;
    std::vecot<Label>                    labels;

When I store a large amount of data, the speed is a bit slow when I access imgs, ids or labels of a certain index on the python side, especially imgs and labels.
For example, I try to call like this

data = Dataset()

import time
st = time.time()
img_5000 = data.imgs[5000]

for i in data.imgs:
    img = i
print("total time is ", time.time() - st)

I tried to use the next code in C++ to speed

 PYBIND11_MAKE_OPAQUE(T) 

py::class_<std::vector<int>>(m, "IntVector")
    .def(py::init<>())
    .def("clear", &std::vector<int>::clear)
    .def("pop_back", &std::vector<int>::pop_back)
    .def("__len__", [](const std::vector<int> &v) { return v.size(); })
    .def("__iter__", [](std::vector<int> &v) {
       return py::make_iterator(v.begin(), v.end());
    }, py::keep_alive<0, 1>()) /* Keep vector alive while iterator is used */
    // ....

and also try to use next code to speed

  py::array_t<int> get_ids(const std::vector<int> &v) {
      py::array_t<int> result(v.size());
      std::memcpy(result.mutable_data(), v.data(), v.size() * sizeof(int));
      return result;
  }

and this code

  int get_ids(const std::vector<int> &v, size_t index) {
      if (index < v.size()) {
          return v[index];
      } else {
          throw std::out_of_range("Index out of range");
      }
  }

However, the access speed is higher, not as fast as the original.

But in fact, when the amount of data is very large, such as 100,000 or 200,000, accessing the original std::vector on the C++ side through index is slower, so I would like to ask if there is anything I can do Any way to optimize speed?