General steps on custom-training a YOLOv8 model

1. Annotate Images:

• Annotate dataset with bounding boxes around objects of interest. Tools like RoboFlow or CVAT can be used for this.


2. Split Data:

• Divide dataset into training and validation sets. The training set is used to train the model, while the validation set is used to evaluate its performance during training.


3. Pre-process the images:

• Standard image pre-processing steps are resizing images to the same size to ensure consistency. Default value used for yolov8 training is 640x640.

4. Augment Images (if needed):

• Data augmentation techniques can be applied to artificially increase the diversity
of your dataset. This helps the model generalize better.


5. Select YOLOv8 model variant:

• A model with more parameters will generally perform better in terms of accuracy
but comes at the cost of slower inference speed.


6. Train the model: (NOTE: model trained using yolov8's default hyperparameters)

• Fine-tune the YOLOv8 model on the annotated dataset using the training set. Monitor metrics such as loss and mAP during training.


7. Validate the Model:

• Evaluate the trained model on the validation set to ensure it generalizes well
to new, unseen data.


8. Inference on Test Set:

• After training, perform inference on a separate test set to assess the model's performance on completely unseen data.