End-cloud Collaborative Object Classification Optimization Method Based on YOLOv8 and Microcontroller K230

Image segmentation and object classification are fundamental tasks in computer vision, widely used in areas such as smart homes and industrial automation. Within the context of IoT and edge computing, embedded devices face significant constraints due to limited computing resources, posing challenges for the deployment of deep learning models. This paper introduces an optimization method for object classification using the K230 microcontroller. By integrating cloud-based DeepLabV3 semantic segmentation with the YOLOv8n object classification model deployed on edge devices, the accuracy of classification tasks is enhanced. Simultaneously, computationally intensive segmentation tasks are offloaded to the cloud, allowing edge devices to focus solely on object classification, thereby achieving an optimal balance between computational efficiency and performance. Experimental results show that this approach significantly improves object classification accuracy, especially in scenarios with complex backgrounds.