fast ways to create a tensor

[maaft]

currently I'm creating a tensor with this approach (as seen in examples):

let data = ndarray::Array::from_shape_fn(
            (1, 3, image.height() as usize, image.width() as usize),
            |(_, c, j, i)| {
                let pixel = image.get_pixel(i as u32, j as u32);
                let channels = pixel.channels();

                // range [0, 255] -> range [0, 1]
                (channels[c] as f32) / 255.0
            },
        );
        
let inputs = ort::inputs!["image" => data].unwrap();

This is super slow and scales very badly for large data.

Are there any parallelization techniques i can use?

I didn't find anything within the ndarray crate.

[audioXD]

You mean like this?

let image = image::RgbImage::new(100, 500);
let ndarray_image: ndarray::Array3<f32> = ndarray::Array3::from_shape_vec(
    (image.height() as usize, image.width() as usize, 3),
    image.into_raw(),
)?.mapv(|v| v as f32 / 255.0);