precomputed.rs

This repository contains a Rust library for downloading, uploading, and manipulating datasets in the Precomputed format popularized by CloudVolume, Neuroglancer and TensorStore.

Features

Feature	Description	Status
Download info	Download and parse the info manifest files	✅
Download chunks	Download data chunks from a Precomputed dataset	✅
Upload chunks	Upload data chunks to a Precomputed dataset	🚧
Create datasets	Create new Precomputed datasets with specified parameters	🚧
Multiscale	Handle multiscale datasets	🚧
Compression	Support for compressed data chunks (e.g., gzip)	🚧
Authentication	Support for authenticated access to private datasets	🚧

Usage

use precomputed_rs::{PrecomputedVolume, Parallelism};
use ndarray::s;

let uri = "https://bossdb-open-data.s3.amazonaws.com/witvliet2020/Dataset_8/em";
// or use `let uri = "file:///path/to/data";` for local filesystem access
let mip = (64.0, 64.0, 30.0);

let cv = PrecomputedVolume::new_at_mip(uri, mip)?
    .fill_missing(0.0)
    .parallel(Parallelism::Auto);

let data = cv.cutout(s![0..512, 0..512, 0..16])?;

println!("dtype  = {:?}", data.dtype());
println!("shape  = {:?}", data.shape());