ConvSequential-SLAM: A Sequence-Based, Training-Less Visual Place Recognition Technique for Changing Environments

Visual Place Recognition (VPR) is the ability to correctly recall a previously visited place under changing viewpoints and appearances. A large number of handcrafted deep-learning-based VPR techniques exist, where former suffer from appearance changes latter have significant computational needs. In this paper, we present new technique, namely ConvSequential-SLAM, that achieves state-of-the-art matching performance challenging conditions. We utilise sequential information block-normalisation handle changes, while using regional-convolutional achieve viewpoint-invariance. analyse content-overlap in-between query frames find minimum sequence length, also re-using image entropy for environment-based length tuning. State-of-the-art reported in contrast 8 contemporary on 4 public datasets. Qualitative insights an ablation study are provided.