Last update - layout detection now works right out of the box with the docstructbench1024 onnx model, but I can't get OCR to work with SVTR/DB, or TROCR, with this repo, so I stopped updating it.

This repo is being used for document layout processing, because of its excellent Rust support for ONNX, to include Paddle and Yolo models. Do not use this repo if you want all the features of the main branch, because this does not have them, and will keep diverging from the main as more changes are made. If you are doing layout processing and/or OCR you might find the finished product useful.

usls is a Rust library integrated with ONNXRuntime, offering a suite of advanced models for Computer Vision and Vision-Language tasks, including:

• YOLO Models: YOLOv5, YOLOv6, YOLOv7, YOLOv8, YOLOv9, YOLOv10, YOLO11, YOLOv12

• OCR Models: FAST, DB(PaddleOCR-Det), SVTR(PaddleOCR-Rec), SLANet, TrOCR, DocLayout-YOLO

By default, none of the following features are enabled. You can enable them as needed:

• auto: Automatically downloads prebuilt ONNXRuntime binaries from Pyke’s CDN for supported platforms.

  • If disabled, you'll need to compile ONNXRuntime from source or download a precompiled package, and then link it manually.

    👉 For Linux or macOS Users

    • Download from the Releases page.
    • Set up the library path by exporting the ORT_DYLIB_PATH environment variable:

      export ORT_DYLIB_PATH=/path/to/onnxruntime/lib/libonnxruntime.so.1.20.1

• ffmpeg: Adds support for video streams, real-time frame visualization, and video export.

  • Powered by video-rs and minifb. For any issues related to ffmpeg features, please refer to the issues of these two crates.

• cuda: Enables the NVIDIA TensorRT provider.

• trt: Enables the NVIDIA TensorRT provider.

• mps: Enables the Apple CoreML provider.

• Using CUDA

  cargo run -r -F cuda --example yolo -- --device cuda:0
  

• Using Apple CoreML

  cargo run -r -F mps --example yolo -- --device mps
  

• Using TensorRT

  cargo run -r -F trt --example yolo -- --device trt
  

• Using CPU

  cargo run -r --example yolo
  

Add usls as a dependency to your project's Cargo.toml

cargo add usls -F cuda

Or use a specific commit:

[dependencies]
usls = { git = "https://github.com/nazeling/usls-doc", rev = "commit-sha" }

This project is licensed under LICENSE.