Questions about incremental training about ultralytics HOT 3 CLOSED

@AntyRia hi there! 😊

Thank you for your kind words and for using our project!

Regarding your question about incremental training, it's crucial to maintain consistency in class labels between your initial and subsequent training sessions. Specifically:

1. Class Order Consistency: The order of the classes in your train_classes.txt should remain the same as in your initial training. This ensures that the model correctly interprets the class indices.

2. YAML File Alignment: Ensure that the train_classes.txt aligns with the class definitions in your corresponding YAML file. The indices (0, 1, 2, etc.) should consistently map to the same classes (person, head, car, etc.).

Here's a quick example to illustrate:

Initial Training:

• train_classes.txt:

  person
  head
  car
  

• YAML file:

  names: [person, head, car]

Incremental Training:

• train_classes.txt:

  person
  head
  car
  

• YAML file:

  names: [person, head, car]

Maintaining this consistency ensures that the model continues to learn correctly without any misinterpretation of class labels.

If you encounter any issues or have further questions, please feel free to share a minimum reproducible code example. This will help us better understand and address your problem. You can refer to our guide on creating a minimum reproducible example here.

Lastly, please ensure you are using the latest versions of torch and ultralytics to avoid any compatibility issues.

Happy training! 🚀

from ultralytics.

Thank you for your reply, this is very important to me!

from ultralytics.

@AntyRia you're welcome! 😊

To ensure we can assist you effectively, could you please provide a minimum reproducible code example? This will help us understand the issue better and work towards a solution. You can find guidance on creating one here.

Additionally, please make sure you're using the latest versions of torch and ultralytics. If not, upgrading might resolve the issue.

Looking forward to your response! 🚀

from ultralytics.