YOLOVAL: local game-image checks with ONNX Runtime CPU (paid model showcase)

[bbshdh56]

I train game-specific object detection models and use ONNX Runtime for local image checks. In a recent small test, I ran a Delta Force model at 640×640 and an Arena Breakout model at 320×320 using CPUExecutionProvider.

For those checks, I compared full images with independently inferred center crops, using a confidence threshold of 0.25 and class-aware NMS at IoU 0.45. The pipeline maps detections back to the original image coordinates for inspection. These are qualitative image checks, not a performance benchmark.

The image below is a broader 32-example overview of my model work; it is separate from that small ONNX Runtime test.

Model catalog — yoloval.cc.cd · Patreon shop — models, prices and downloads

Open the full-resolution image · GitHub overview page

Creator disclosure: I train and sell these models. The links above lead to my paid model catalog and shop.