Evaluating 3- D Features for Model Matching

This paper evaluates the performance of various al-gorithms in leveraging 3-D features for model matching. However, the lack of a common evaluation framework makes it challenging to compare different choices in the processing pipeline. In this paper, I introduce Proctor, an open-source framework designed for comparing 3-D features in model matching tasks. Proctor aims to foster easier sharing of ideas and methods by providing a modular architecture for evaluating various 3-D features. In-spired by successful standardization efforts in robotics and com-puter vision, Proctor offers a common platform for researchers to assess the performance of 3-D perception algorithms. I present the motivations behind Proctor and describe its modular archi-tecture, encouraging a community of users and contributors. The framework facilitates the comparison of different 3-D features, supporting tasks such as object recognition, pose estimation, and detection in clean or cluttered scenes. my evaluation includes a diverse range of features and datasets, demonstrating the effectiveness of Proctor in assessing the accuracy and speed of 3- D perception algorithms. Through Proctor, I aim to accelerate progress in 3-D perception research by providing a standardized platform for evaluating and comparing diverse approaches.