Darknet v1.1.1

See also new iChess.io v12.27!!!

📄 This work is described in the scientific paper: https://doi.org/10.1002/cpe.6936

📦 Platform Support

• ✅ macOS (Intel / Apple Silicon)
• ✅ Ubuntu Linux 20.04+
• ⚠️ Windows 10/11 (experimental OpenCL build)

Windows Build Guide:

https://iblog.isowa.io/2021/11/20/darknet-on-opencl-on-windows-11-x64

🧠 Training & Optimization Tips

• Run from RAMDisk to reduce disk wear and speed up training/inference:

  • Linux: sudo mount -t tmpfs -o size=4096M tmpfs /your/ramdisk
  • macOS: diskutil erasevolume HFS+ "ramdisk" $(hdiutil attach -nomount ram://8388608)

• Replace clBLAS with CLBlast for improved GEMM performance:

git apply patches/clblast.patch

🔗 Related Projects

• Darknet-vNext — CUDA-enhanced variant
• YOLO on OpenCL

📽️ Demos & Videos

• 4 x AMD Radeon RX 6900 XT Demo (macOS)
• YOLO2 Multi-GPU Result
• YOLO Training

🙏 Acknowledgements

Created by Piotr Sowa — AI researcher, GPU software engineer, and creator of iChess.io. More information and tutorials at iBlog.isowa.io.

---

For citations, academic usage, or collaboration inquiries, feel free to reach out via GitHub or LinkedIn.