YOLO Performance Metrics - Ultralytics YOLOv8 Docs

[giscus[bot]]

YOLO Performance Metrics - Ultralytics YOLOv8 Docs

A comprehensive guide on various performance metrics related to YOLOv8, their significance, and how to interpret them.

https://docs.ultralytics.com/guides/yolo-performance-metrics/

[kholidiyah]

I want to analyze F1-score that get from Yolov8 training, how do i get the value of F1-score and bitrate from training process?

[pderrenger]

@kholidiyah during the training process with YOLOv8, the F1-score is automatically calculated and logged for you. To obtain the F1-score and other metrics such as precision, recall, and mAP (mean Average Precision), you can follow these steps:

1. Ensure that you have validation enabled during training by setting val: True in your training configuration.
2. After each training epoch, the model will be evaluated on the validation dataset, and performance metrics will be calculated.
3. The results, including the F1-score, will be displayed in the terminal and saved to the results.csv file in your run directory (runs/train/exp*/results.csv by default).
4. To analyze the F1-score, you can open the results.csv file and look for the column labeled F1 to see the F1-score for each epoch.

As for the bitrate, if you are referring to the model size or the inference speed, these are not directly referred to as "bitrate" in the context of YOLOv8. However, you can find the model size (in megabytes) as a .pt file in your run directory, and you can benchmark the inference speed using the benchmark mode with the desired settings.

If you need further details on how to interpret these metrics or how to use the benchmark mode, please refer to the Ultralytics documentation.