zune-image VS image

Thanks to the zune-image library, serde-numpy is now supporting PNG and JPEG. Enabling faster io for neural net training loops and other image processing in python

It has been extensively tested on 600,000 real world images, as well as fuzzed in various ways, and is now ready for production use!

I might have misunderstood what you've written, but is there a reason you can't use https://github.com/etemesi254/zune-image as a JPEG decoder? It has minimal use of unsafe and is performant.

It uses SIMD for colorspace conversion and IDCT, the code can be found here and here.

You can see a Github Action that zune-inflate uses here, it's mostly self-explanatory.

I've just added a roundtrip fuzzer to the Rust translation of it so that we can start weeding out the bugs: https://github.com/etemesi254/zune-image/pull/13 Right now it finds some decoding failures in a few seconds.

With the help of the image crate, loading an image from disk is fairly straightforward.

FTR there is a WebP decoder implementation in safe Rust in the image crate: https://github.com/image-rs/image

It used to be quite incomplete for a long time, but work last year has implemented many webp features. Chromium now has a policy of allowing the use of Rust dependencies, so maybe Chromium could start adopting it?

Hello! So I have a local app that I am currently running on desktop (windows). I'm using egui for the UI, and the program basically opens a folder, gets all the images in the folder, and then uses the image-rs library to resize and create a grid of images / some other operations.

png is the de-facto standard Rust crate for reading and writing PNG images, used e.g. by the image crate.

There's also a CLI to convert between formats. It uses the crate image.

exr is a is a 100% Rust and 100% safe code library for reading and writing OpenEXR images. It is used by the popular image crate to read and write OpenEXR.

We're currently looking for contributors to add support for zune-jpeg to the image crate. The image maintainers are open to it, but don't have the capacity to do it themselves. You can find more details here.

It would be helpful to contribute to the general change towards an untyped buffer with runtime representations for color space information and fallible conversions. (https://github.com/image-rs/image/pull/1718). It's currently stalling on having too few eyeballs to judge the impact and need for that complexity.

It uses https://github.com/1Password/arboard and https://github.com/image-rs/image under the hood. I haven't tested deeply but I assume it should handle many formats out of the box.