Repository navigation
YOLOVAL: Arena Breakout ONNX-to-ncnn CPU check + 32 examples (paid models) #7238
bbshdh56
started this conversation in
Show and tell
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
I publish YOLOVAL, a catalog of paid game-specific ONNX detection models. Here is a small ONNX-to-ncnn CPU conversion check alongside the broader 32-example visual gallery.
Full-resolution English overview · Model catalog · Patreon shop — prices and downloads
The overview preserves .onnx filenames and the colorful yoloval.cc.cd watermark. It covers Delta Force, Arena Breakout, CS2, Apex, CrossFire, Call of Duty, THE FINALS, Game for Peace, SKJH, PUBG, Overwatch, Strinova, and red/yellow/purple VALORANT enemies. Available sizes include 192, 256, 320, 416 and 640, depending on the model.
What I checked
Using pnnx 20261020, ncnn 1.0.20261020, and ONNX Runtime 1.24.4, I converted local copies with fp16=0 and compared one previously prepared real-image tensor per model. ncnn used CPU, four threads, and disabled FP16/BF16 storage and packing layout.
The command used for the Arena Breakout model was
pnnx model.onnx inputshape=[1,3,320,320] fp16=0. The generated ncnn weight file was 10,358,060 bytes. Original ONNX files were unchanged.This is a small numerical check, not an accuracy or FPS benchmark. The broader gallery shows ONNX model examples; it does not imply every pictured model has been validated in ncnn. I'm the independent publisher of the paid catalog.
All reactions