lonacn/vision_simple repository overview

⁠🚀 vision-simple 🚀

⁠

A lightweight Cross-Platform Visual Inference Library

The library leverages the ONNXRuntime engine, supporting multiple Execution Providers such as DirectML, CUDA, TensorRT and RKNPU.

该库利用ONNXRuntime引擎，支持多个执行提供器，如DirectML、CUDA、TensorRT和RKNPU。

github⁠

⁠Key Features：

• Cross-Platform: Supports windows/x64, linux/x86_64, linux/arm64 and linux/riscv64.
• Multi-Device Support: Compatible with CPU, GPU, and RKNPU.
• Lightweight: Static builds under 20MB, with YOLO and OCR inference consuming only 300MB of memory.
• Quick Deployment:
  • One-click Compilation: Provides verified build scripts for multiple platforms.
  • Container Deployment: Easily deploy via Docker, Podman, or Container.
  • HTTP API: Offers an HTTP API for non-real-time applications.

⁠Docker Deployment:

docker run -it --rm --name vs -p 11451:11451 lonacn/vision_simple:0.4.0-cpu-x86_64

list models

curl --location --request GET 'http://127.0.0.1:11451/v0/infer/models'

⁠Notice

⁠NPU Inference

By default, CPU inference is used in rknpu-arm64 container,u can edit/app/config/server.yamland set infer_ep=kRKNPU,for example:

options:
  static_path: "assets/static"
  infer_framework: "kONNXRUNTIME"
  infer_ep: "kRKNPU"
  infer_device: "0"