Artem-N

Results 8 comments of Artem-N

> Can I transplant it on the RK3588 board? yes, first you need to export your model to the onnx format ( in this repo script export.py ), then you...

> what did you do? Did you have a solution for this? unfortunately not, I supplemented the classes with images and hope that Yolo7 is smart and will be able...

Hi guys, I tried updating, and after training a custom network once again (which I’ve done multiple times before), I attempted to convert the model to ONNX — and got...

> So when you use the predict method of a different model in this exact code, you don't see high CPU use? How's GPU utilization? Hi, I have a similar...

> The difference lies in the preprocessing approach. YOLO (ultralytics) resizes frames on CPU before transferring to GPU, while RF-DETR transfers to GPU first, then resizes. This means RF-DETR transfers...

> The difference lies in the preprocessing approach. YOLO (ultralytics) resizes frames on CPU before transferring to GPU, while RF-DETR transfers to GPU first, then resizes. This means RF-DETR transfers...

> another possible scenario that causes high cpu usage is if you transfer the image to gpu in float32 instead of uint8. i'm going to test, thanks. upd. After many...