YOLO26 Object Detection

[tbc725]

I'm not sure if this project died, but here is a skeleton of YOLO26 for Windows CPU.

This is basically taking YOLOv8 and modifying it so it works with YOLO26 and CPU only. I did not test the GPU part.

Important

• The user will need to download the YOLO26 models manually from Ultralytics and put them into the "assets" folder.
• The user will need to copy the “bin” folder from the YOLOv8 modules and then update their Ultralytics module to the latest version.
• CPU only. GPU not tested.
• Windows only. Did not try on other platforms.

skeleton-ObjectDetectionYOLO26.zip

[simshaun]

Thanks. If you're a noob like me:

To download the YOLO26 models:

Go to https://huggingface.co/Ultralytics/YOLO26/tree/main
Look for the variants you want. I downloaded the yolo26n.pt, yolo26s.pt, and yolo26m.pt files to try out.

To update the Ultralytics model to the latest version:

Navigate down into your modules folder to somewhere like this:
C:\Program Files\CodeProject\AI\modules\ObjectDetectionYOLO26\bin\windows\python39\venv\Scripts

Open your terminal in that folder, then run .\python.exe -m pip install -U ultralytics

---

I'm testing this out on a Beelink EQ13 with an Intel N200 processor.

YOLO26s, ~500ms to detect.
YOLO26m, ~950ms to detect.
YOLO26l, ~1200ms to detect

Not bad.

[MarkDRenard]

Just an FYI: I have a complete ready to download and install ObjectDetectionYOLO26 module for CodeProject.AI v2.9.5 on my GitHub: https://github.com/MarkDRenard/CodeProject.AI-Modules that I developed for Windows x64 with NVIDIA RTX GPU running Python 3.12, CUDA 12.4 and cuDNN 9 inside the module and a full modern performance array/stack. The module is fully self-contained including automated downloading and post-install verification of the Detection, Segmentation and Custom .pt models from my GitHub YOLO26-Models release.

I am running a stand alone CodeProject.AI v2.9.5 Windows 11, I7-9900 CPU, 32gb RAM, NVIDIA RTX 2060 GPU Server paired with a stand alone Blue Iris Security v6 Windows 11, I7-9900 CPU, 32gb RAM, NVIDIA GTX 1060 GPU Server and using the default YOLO26n.pt model, average 74-82ms to detect on the CodeProject.AI 'Test' photo and 100-150ms(+-) on Blue Iris trigger detects. I will also release a .onnx models version, after testing. If you're interested, please check it out. -mark👍🏼🐾

CodeProject.AI Explorer Test Photo:

Label Confidence

0 chair 86%
1 chair 83%
2 potted plant 72%
3 vase 53%
Processed by ObjectDetectionYOLO26
Processed on localhost
API call round trip 82 ms
Processing 75 ms
Inference 74 ms
Timestamp (UTC) Thu, 15 Oct 2026 20:54:39 GMT

[tbc725]

Thanks a lot for your hard work!