Exploring a Mixer-Based Model for Enhanced Loop Closure Detection in Robotic Navigation
Document Type
Conference Proceeding
Publication Date
2-7-2025
Abstract
Mixers offer a strong balance between resource efficiency and accuracy which makes them promising feature extractors for loop closure detection (LCD). This paper evaluates the performance of Mixers in LCD by comparing them against handcrafted and learned feature extractors. Using a Siamese architecture with dynamic margin distance, Mixers effectively distinguish loops from non-loops. In most of the chosen datasets, the proposed model achieved higher MR scores than that of the other models while requiring fewer resources. These results demonstrate that the Mixer-based approach is both effective and resource-efficient compared to current methods.
Recommended Citation
Kendra Kirsten L. Go and Raphael B. Alampay. 2025. Exploring a Mixer-Based Model for Enhanced Loop Closure Detection in Robotic Navigation. In Proceedings of the 2024 7th International Conference on Computational Intelligence and Intelligent Systems (CIIS '24). Association for Computing Machinery, New York, NY, USA, 119–124. https://doi.org/10.1145/3708778.3708795
