Implementation of the VAE in Rust 🦀 + Burn 🔥.
I collected some images of characters in Detective Conan, then converted them to sketch using Anime2Sketch
Please note that the customized dataset will not be made public due to copyright issues, this repo only works with MNIST dataset.
Make sure all the folders below exist
.
├── Cargo.lock
├── Cargo.toml
├── artifacts
├── images
├── readme.md
├── src
├── target
Then run cargo run --release
Target and reconstruction
Sampling results
As a comparison, here are some results from a VAE implemented in pytorch
Sketch to image results from windy AI