Convert torchvision functions to libtorch C++ codes

C++
1 
from torchvision.tv_tensors import BoundingBoxes
from torchvision.tv_tensors import Mask
...
        self._pred_bboxes = BoundingBoxes(
            model_output[0]["boxes"][scores_mask]
            format="xyxy",
            canvas_size=self._input_img.size[::-1],
        )

        self._pred_masks = torch.nn.functional.interpolate(
            model_output[0]["masks"][scores_mask], size=self._color_image.size[::-1]
        )

        # mask is a two classificaiton problem
        self._pred_masks = torch.concat(
            [
                Mask(torch.where(mask >= 0.5, 1, 0), dtype=torch.bool)
                for mask in self._pred_masks
            ]
        )

I want to transform it into C++ codes.

I’ve found function for interpolate, it is as below

        torch::Tensor _pred_masks = torch::nn::functional::interpolate(
                image_tensor,
                F::InterpolateFuncOptions()
                        .size(std::vector<int64_t>({_color_img.rows, _color_img.cols}))
        );

But for torchvision, I havent found corresponding functions. I decided to store the _pred_bboxes as torch::tensor or std::vector.

For torchvision::tv_tensors::Mask, I don’t even know what it means. the documentation says

torch.Tensor subclass for segmentation and detection masks

but it is very vague for what it does.

I have read this post Is torchvision also available in C++? - #6 by Shisho_Sama but didnt find useful information about my issue.