To achieve precise tracking of the continuum arm’s shape and movements, the YOLOv8 model was trained on the synthetic dataset. The training process involved annotating the synthetic images with specific labels for the continuum arm’s segments, enabling the model to learn the distinct shapes and orientations of the arm. Using YOLOv8’s advanced object detection capabilities, the model performed real-time segmentation of the continuum arm, accurately identifying its shape and position across various simulated scenarios.

This high level of accuracy was achieved through iterative training, adjustments in hyperparameters, and data augmentation techniques. The precise segmentation capability was crucial for enabling accurate control over the arm's movements, facilitating detailed analysis of each position and directly informing the fine-tuning of control algorithms.