Advanced Customization

Both the Ultralytics YOLO command-line and Python interfaces are simply a high-level abstraction on the base engine executors. Let's take a look at the Trainer engine.

Watch: Mastering Ultralytics YOLOv8: Advanced Customization

BaseTrainer

BaseTrainer contains the generic boilerplate training routine. It can be customized for any task based over overriding the required functions or operations as long the as correct formats are followed. For example, you can support your own custom model and dataloader by just overriding these functions:

• get_model(cfg, weights) - The function that builds the model to be trained
• get_dataloader() - The function that builds the dataloader More details and source code can be found in BaseTrainer Reference

DetectionTrainer

Here's how you can use the YOLOv8 DetectionTrainer and customize it.

from ultralytics.models.yolo.detect import DetectionTrainer

trainer = DetectionTrainer(overrides={...})
trainer.train()
trained_model = trainer.best  # get best model

Customizing the DetectionTrainer

Let's customize the trainer to train a custom detection model that is not supported directly. You can do this by simply overloading the existing the get_model functionality:

from ultralytics.models.yolo.detect import DetectionTrainer


class CustomTrainer(DetectionTrainer):
    def get_model(self, cfg, weights):
        """Loads a custom detection model given configuration and weight files."""
        ...


trainer = CustomTrainer(overrides={...})
trainer.train()

You now realize that you need to customize the trainer further to:

• Customize the loss function.
• Add callback that uploads model to your Google Drive after every 10 epochs Here's how you can do it:

from ultralytics.models.yolo.detect import DetectionTrainer
from ultralytics.nn.tasks import DetectionModel


class MyCustomModel(DetectionModel):
    def init_criterion(self):
        """Initializes the loss function and adds a callback for uploading the model to Google Drive every 10 epochs."""
        ...


class CustomTrainer(DetectionTrainer):
    def get_model(self, cfg, weights):
        """Returns a customized detection model instance configured with specified config and weights."""
        return MyCustomModel(...)


# callback to upload model weights
def log_model(trainer):
    """Logs the path of the last model weight used by the trainer."""
    last_weight_path = trainer.last
    print(last_weight_path)


trainer = CustomTrainer(overrides={...})
trainer.add_callback("on_train_epoch_end", log_model)  # Adds to existing callback
trainer.train()

To know more about Callback triggering events and entry point, checkout our Callbacks Guide

Other engine components

There are other components that can be customized similarly like Validators and Predictors. See Reference section for more information on these.

FAQ

How do I customize the Ultralytics YOLOv8 DetectionTrainer for specific tasks?

To customize the Ultralytics YOLOv8 DetectionTrainer for a specific task, you can override its methods to adapt to your custom model and dataloader. Start by inheriting from DetectionTrainer and then redefine methods like get_model to implement your custom functionalities. Here's an example:

from ultralytics.models.yolo.detect import DetectionTrainer


class CustomTrainer(DetectionTrainer):
    def get_model(self, cfg, weights):
        """Loads a custom detection model given configuration and weight files."""
        ...


trainer = CustomTrainer(overrides={...})
trainer.train()
trained_model = trainer.best  # get best model

For further customization like changing the loss function or adding a callback, you can reference our Callbacks Guide.

What are the key components of the BaseTrainer in Ultralytics YOLOv8?

The BaseTrainer in Ultralytics YOLOv8 serves as the foundation for training routines and can be customized for various tasks by overriding its generic methods. Key components include:

• get_model(cfg, weights) to build the model to be trained.
• get_dataloader() to build the dataloader.

For more details on the customization and source code, see the BaseTrainer Reference.

How can I add a callback to the Ultralytics YOLOv8 DetectionTrainer?

You can add callbacks to monitor and modify the training process in Ultralytics YOLOv8 DetectionTrainer. For instance, here's how you can add a callback to log model weights after every training epoch:

from ultralytics.models.yolo.detect import DetectionTrainer


# callback to upload model weights
def log_model(trainer):
    """Logs the path of the last model weight used by the trainer."""
    last_weight_path = trainer.last
    print(last_weight_path)


trainer = DetectionTrainer(overrides={...})
trainer.add_callback("on_train_epoch_end", log_model)  # Adds to existing callbacks
trainer.train()

For further details on callback events and entry points, refer to our Callbacks Guide.

Why should I use Ultralytics YOLOv8 for model training?

Ultralytics YOLOv8 offers a high-level abstraction on powerful engine executors, making it ideal for rapid development and customization. Key benefits include:

• Ease of Use: Both command-line and Python interfaces simplify complex tasks.
• Performance: Optimized for real-time object detection and various vision AI applications.
• Customization: Easily extendable for custom models, loss functions, and dataloaders.

Learn more about YOLOv8's capabilities by visiting Ultralytics YOLO.

Can I use the Ultralytics YOLOv8 DetectionTrainer for non-standard models?

Yes, Ultralytics YOLOv8 DetectionTrainer is highly flexible and can be customized for non-standard models. By inheriting from DetectionTrainer, you can overload different methods to support your specific model's needs. Here's a simple example:

from ultralytics.models.yolo.detect import DetectionTrainer


class CustomDetectionTrainer(DetectionTrainer):
    def get_model(self, cfg, weights):
        """Loads a custom detection model."""
        ...


trainer = CustomDetectionTrainer(overrides={...})
trainer.train()

For more comprehensive instructions and examples, review the DetectionTrainer documentation.

Comments