Robotic kiwifruit harvesting systems are currently being introduced to improve the reliability and farming yields of kiwifruit harvesting operations. Machine learning is widely used to carry out the visual detection tasks required of such systems. This paper specifically compares two types of machine learning algorithms: the multivariate alternating decision tree and deep learning based kiwifruit classifiers. The purpose of the study is to investigate the cost of implementation against the classification performance. Thus, discussion is centred around computational cost and its impacts on the overall system architecture. We found that the traditional decision tree classifiers can achieve comparable classification performance at a fraction of the cost and complexity, providing robust and cost-effective instrument design.