Ayush Chaurasia
Ayush Chaurasia
@ercanhocaninyeri running it 5 times in a loop? Or running the file itself 5 times? Can you please provide the exact steps and code to reproduce?
Yeah, I see the spreadsheet. I don't understand the latency problem but time that you see in predict is calculated using time module itself
Yeah but source setup should only be done in the 1st call. and pre/post processing is done with val too right? But I think when comparing train, val it'll come...
For cli: ``` yolo... verbose=False ``` for python it is false by default.
@ch-hristov thanks for asking. `model.train()` accepts all training arguments mentioned here https://docs.ultralytics.com/config/#training for example: `` mode.train(data=..., epochs=1, lr0=0.001..,..) ```
Yes! mutli-scale training has not been added yet. I'll put it in our TODOs.
See this official example for YOLOv8 ONNX opencv python inference https://github.com/ultralytics/ultralytics/tree/main/examples/YOLOv8-OpenCV-ONNX-Python
Pass verbose=False in the inference function
Then it's a bug. The masks should be drawn on verbose=False. I'll fix it tomorrow