Kanji Tanaka

Shogo Hanada, Kanji Tanaka

In this paper, we explore the challenging 1-to-N map matching problem, which exploits a compact description of map data, to improve the scalability of map matching techniques used by various robot vision tasks. We propose a first method explicitly aimed at fast succinct map matching, which consists only of map-matching subtasks. These tasks include offline map matching attempts to find a compact part-based scene model that effectively explains each map using fewer larger parts. The tasks also include an online map matching attempt to efficiently find correspondence between the part-based maps. Our part-based scene modeling approach is unsupervised and uses common pattern discovery (CPD) between the input and known reference maps. This enables a robot to learn a compact map model without human intervention. We also present a practical implementation that uses the state-of-the-art CPD technique of randomized visual phrases (RVP) with a compact bounding box (BB) based part descriptor, which consists of keypoint and descriptor BBs. The results of our challenging map-matching experiments, which use a publicly available radish dataset, show that the proposed approach achieves successful map matching with significant speedup and a compact description of map data that is tens of times more compact. Although this paper focuses on the standard 2D point-set map and the BB-based part representation, we believe our approach is sufficiently general to be applicable to a broad range of map formats, such as the 3D point cloud map, as well as to general bounding volumes and other compact part representations.

Yuki Minase, Kanji Tanaka

Robust person tracking is a critical capability for autonomous mobile robots operating in diverse and unpredictable environments. While RGB-D tracking has shown high precision, its performance severely degrades under challenging illumination conditions, such as total darkness or intense backlighting. To achieve all-weather robustness, this paper proposes a novel Thermal-Infrared and Depth (TIR-D) tracking architecture that leverages the standard sensor suite of SLAM-capable robots, namely LiDAR and TIR cameras. A major challenge in TIR-D tracking is the scarcity of annotated multi-modal datasets. To address this, we introduce a sequential knowledge transfer strategy that evolves structural priors from a large-scale thermal-trained model into the TIR-D domain. By employing a differential learning rate strategy -- referred to as ``Fine-grained Differential Learning Rate Strategy'' -- we effectively preserve pre-trained feature extraction capabilities while enabling rapid adaptation to geometric depth cues. Experimental results demonstrate that our proposed TIR-D tracker achieves superior performance, with an Average Overlap (AO) of 0.700 and a Success Rate (SR) of 58.7\%, significantly outperforming conventional RGB-transfer and single-modality baselines. Our approach provides a practical and resource-efficient solution for robust human-following in all-weather robotics applications.

Koji Takeda, Kanji Tanaka, Yoshimasa Nakamura

The problem of image change detection via everyday indoor robot navigation is explored from a novel perspective of the self-attention technique. Detecting semantically non-distinctive and visually small changes remains a key challenge in the robotics community. Intuitively, these small non-distinctive changes may be better handled by the recent paradigm of the attention mechanism, which is the basic idea of this work. However, existing self-attention models require significant retraining cost per domain, so it is not directly applicable to robotics applications. We propose a new self-attention technique with an ability of unsupervised on-the-fly domain adaptation, which introduces an attention mask into the intermediate layer of an image change detection model, without modifying the input and output layers of the model. Experiments, in which an indoor robot aims to detect visually small changes in everyday navigation, demonstrate that our attention technique significantly boosts the state-of-the-art image change detection model.

Kanya Kurauchi, Kanji Tanaka, Ryogo Yamamoto et al.

The training of a next-best-view (NBV) planner for visual place recognition (VPR) is a fundamentally important task in autonomous robot navigation, for which a typical approach is the use of visual experiences that are collected in the target domain as training data. However, the collection of a wide variety of visual experiences in everyday navigation is costly and prohibitive for real-time robotic applications. We address this issue by employing a novel {\it domain-invariant} NBV planner. A standard VPR subsystem based on a convolutional neural network (CNN) is assumed to be available, and its domain-invariant state recognition ability is proposed to be transferred to train the domain-invariant NBV planner. Specifically, we divide the visual cues that are available from the CNN model into two types: the output layer cue (OLC) and intermediate layer cue (ILC). The OLC is available at the output layer of the CNN model and aims to estimate the state of the robot (e.g., the robot viewpoint) with respect to the world-centric view coordinate system. The ILC is available within the middle layers of the CNN model as a high-level description of the visual content (e.g., a saliency image) with respect to the ego-centric view. In our framework, the ILC and OLC are mapped to a state vector and subsequently used to train a multiview NBV planner via deep reinforcement learning. Experiments using the public NCLT dataset validate the effectiveness of the proposed method.

Haruki Nakata, Kanji Tanaka, Koji Takeda

In robot vision, self-attention has recently emerged as a technique for capturing non-local contexts. In this study, we introduced a self-attention mechanism into the intersection recognition system as a method to capture the non-local contexts behind the scenes. An intersection classification system comprises two distinctive modules: (a) a first-person vision (FPV) module, which uses a short egocentric view sequence as the intersection is passed, and (b) a third-person vision (TPV) module, which uses a single view immediately before entering the intersection. The self-attention mechanism is effective in the TPV module because most parts of the local pattern (e.g., road edges, buildings, and sky) are similar to each other, and thus the use of a non-local context (e.g., the angle between two diagonal corners around an intersection) would be effective. This study makes three major contributions. First, we proposed a self-attention-based approach for intersection classification using TPVs. Second, we presented a practical system in which a self-attention-based TPV module is combined with an FPV module to improve the overall recognition performance. Finally, experiments using the public KITTI dataset show that the above self-attention-based system outperforms conventional recognition based on local patterns and recognition based on convolution operations.

Ryogo Yamamoto, Kanji Tanaka

The use of relative attribute (e.g., beautiful, safe, convenient) -based image embeddings in visual place recognition, as a domain-adaptive compact image descriptor that is orthogonal to the typical approach of absolute attribute (e.g., color, shape, texture) -based image embeddings, is explored in this paper.

Ryogo Yamamoto, Kanji Tanaka

Cross-view self-localization is a challenging scenario of visual place recognition in which database images are provided from sparse viewpoints. Recently, an approach for synthesizing database images from unseen viewpoints using NeRF (Neural Radiance Fields) technology has emerged with impressive performance. However, synthesized images provided by these techniques are often of lower quality than the original images, and furthermore they significantly increase the storage cost of the database. In this study, we explore a new hybrid scene model that combines the advantages of view-invariant appearance features computed from raw images and view-dependent spatial-semantic features computed from synthesized images. These two types of features are then fused into scene graphs, and compressively learned and recognized by a graph neural network. The effectiveness of the proposed method was verified using a novel cross-view self-localization dataset with many unseen views generated using a photorealistic Habitat simulator.

Jonathan Tay Yu Liang, Kanji Tanaka

The emerging ``Floor plan from human trails (PfH)" technique has great potential for improving indoor robot navigation by predicting the traversability of occluded floors. This study presents an innovative approach that replaces first-person-view sensors with a third-person-view monocular camera mounted on the observer robot. This approach can gather measurements from multiple humans, expanding its range of applications. The key idea is to use two types of trackers, SLAM and MOT, to monitor stationary objects and moving humans and assess their interactions. This method achieves stable predictions of traversability even in challenging visual scenarios, such as occlusions, nonlinear perspectives, depth uncertainty, and intersections involving multiple humans. Additionally, we extend map quality metrics to apply to traversability maps, facilitating future research. We validate our proposed method through fusion and comparison with established techniques.

Mitsuaki Uno, Kanji Tanaka, Daiki Iwata et al.

Object Goal Navigation (OGN) is a fundamental task for robots and AI, with key applications such as mobile robot image databases (MRID). In particular, mapless OGN is essential in scenarios involving unknown or dynamic environments. This study aims to enhance recent modular mapless OGN systems by leveraging the commonsense reasoning capabilities of large language models (LLMs). Specifically, we address the challenge of determining the visiting order in frontier-based exploration by framing it as a frontier ranking problem. Our approach is grounded in recent findings that, while LLMs cannot determine the absolute value of a frontier, they excel at evaluating the relative value between multiple frontiers viewed within a single image using the view image as context. We dynamically manage the frontier list by adding and removing elements, using an LLM as a ranking model. The ranking results are represented as reciprocal rank vectors, which are ideal for multi-view, multi-query information fusion. We validate the effectiveness of our method through evaluations in Habitat-Sim.

Yudai Noda, Kanji Tanaka

Object-Goal Navigation (ObjectNav) requires an agent to find and navigate to a target object category in unknown environments. While recent Large Language Model (LLM)-based agents exhibit zero-shot reasoning, they often rely on a "reactive" paradigm that lacks explicit spatial memory, leading to redundant exploration and myopic behaviors. To address these limitations, we propose a transition from reactive AI to "Map-Based AI" by integrating LLM-based semantic inference with a hybrid topological-grid mapping system. Our framework employs a fine-tuned Llama-2 model via Low-Rank Adaptation (LoRA) to infer semantic zone categories and target existence probabilities from verbalized object observations. In this study, a "zone" is defined as a functional area described by the set of observed objects, providing crucial semantic co-occurrence cues for finding the target. This semantic information is integrated into a topological graph, enabling the agent to prioritize high-probability areas and perform systematic exploration via Traveling Salesman Problem (TSP) optimization. Evaluations in the AI2-THOR simulator demonstrate that our approach significantly outperforms traditional frontier exploration and reactive LLM baselines, achieving a superior Success Rate (SR) and Success weighted by Path Length (SPL).

Dingrui Li, Dedi Guo, Kanji Tanaka

In 3D LiDAR-based robot self-localization, pole-like landmarks are gaining popularity as lightweight and discriminative landmarks. This work introduces a novel approach called "discriminative rotation-invariant poles," which enhances the discriminability of pole-like landmarks while maintaining their lightweight nature. Unlike conventional methods that model a pole landmark as a 3D line segment perpendicular to the ground, we propose a simple yet powerful approach that includes not only the line segment's main body but also its surrounding local region of interest (ROI) as part of the pole landmark. Specifically, we describe the appearance, geometry, and semantic features within this ROI to improve the discriminability of the pole landmark. Since such pole landmarks are no longer rotation-invariant, we introduce a novel rotation-invariant convolutional neural network that automatically and efficiently extracts rotation-invariant features from input point clouds for recognition. Furthermore, we train a pole dictionary through unsupervised learning and use it to compress poles into compact pole words, thereby significantly reducing real-time costs while maintaining optimal self-localization performance. Monte Carlo localization experiments using publicly available NCLT dataset demonstrate that the proposed method improves a state-of-the-art pole-based localization framework.

Kenta Tsukahara, Kanji Tanaka

A typical assumption in state-of-the-art self-localization models is that an annotated training dataset is available for the target workspace. However, this is not necessarily true when a robot travels around the general open world. This work introduces a novel training scheme for open-world distributed robot systems. In our scheme, a robot (``student") can ask the other robots it meets at unfamiliar places (``teachers") for guidance. Specifically, a pseudo-training dataset is reconstructed from the teacher model and then used for continual learning of the student model under domain, class, and vocabulary incremental setup. Unlike typical knowledge transfer schemes, our scheme introduces only minimal assumptions on the teacher model, so that it can handle various types of open-set teachers, including those uncooperative, untrainable (e.g., image retrieval engines), or black-box teachers (i.e., data privacy). In this paper, we investigate a ranking function as an instance of such generic models, using a challenging data-free recursive distillation scenario, where a student once trained can recursively join the next-generation open teacher set.

Sun Yifei, Li Dingrui, Ye Minying et al.

In 3D point cloud-based visual self-localization, pole landmarks have a great potential as landmarks for accurate and reliable localization due to their long-term stability under seasonal and weather changes. In this study, we aim to explore the use of recently developed deep learning models for pole classification in the context of pole landmark-based self-localization. Specifically, the proposed scheme consists of two main modules: pole map matching and pole class matching. In the former module, local pole map is constructed and its configuration is compared against a precomputed global pole map. An efficient RANSAC map matching is employed to achieve a good tradeoff between computational efficiency and accuracy. In the latter pole class matching module, the local and global poles paired by the RANSAC map-matching are further compared by means of pole attribute class. To this end, a predefined set of pseudo pole classes is learned via k-means clustering in a self-supervised manner. Experiments using publicly available NCLT dataset showed that the pole-like landmark classification method has an improved effect on the visual self-localization system compared with the baseline method.

Tomoya Iwasaki, Kanji Tanaka, Kenta Tsukahara

Visual place classification from a first-person-view monocular RGB image is a fundamental problem in long-term robot navigation. A difficulty arises from the fact that RGB image classifiers are often vulnerable to spatial and appearance changes and degrade due to domain shifts, such as seasonal, weather, and lighting differences. To address this issue, multi-sensor fusion approaches combining RGB and depth (D) (e.g., LIDAR, radar, stereo) have gained popularity in recent years. Inspired by these efforts in multimodal RGB-D fusion, we explore the use of pseudo-depth measurements from recently-developed techniques of ``domain invariant" monocular depth estimation as an additional pseudo depth modality, by reformulating the single-modal RGB image classification task as a pseudo multi-modal RGB-D classification problem. Specifically, a practical, fully self-supervised framework for training, appropriately processing, fusing, and classifying these two modalities, RGB and pseudo-D, is described. Experiments on challenging cross-domain scenarios using public NCLT datasets validate effectiveness of the proposed framework.

Mitsuki Yoshida, Kanji Tanaka, Ryogo Yamamoto et al.

Semantic localization, i.e., robot self-localization with semantic image modality, is critical in recently emerging embodied AI applications (e.g., point-goal navigation, object-goal navigation, vision language navigation) and topological mapping applications (e.g., graph neural SLAM, ego-centric topological map). However, most existing works on semantic localization focus on passive vision tasks without viewpoint planning, or rely on additional rich modalities (e.g., depth measurements). Thus, the problem is largely unsolved. In this work, we explore a lightweight, entirely CPU-based, domain-adaptive semantic localization framework, called graph neural localizer. Our approach is inspired by two recently emerging technologies: (1) Scene graph, which combines the viewpoint- and appearance- invariance of local and global features; (2) Graph neural network, which enables direct learning/recognition of graph data (i.e., non-vector data). Specifically, a graph convolutional neural network is first trained as a scene graph classifier for passive vision, and then its knowledge is transferred to a reinforcement-learning planner for active vision. Experiments on two scenarios, self-supervised learning and unsupervised domain adaptation, using a photo-realistic Habitat simulator validate the effectiveness of the proposed method.

Kyosuke Tashiro, Koji Takeda, Kanji Tanaka et al.

Visual defect detection (VDD) for high-mix low-volume production of non-convex metal objects, such as high-pressure cylindrical piping joint parts (VDD-HPPPs), is challenging because subtle difference in domain (e.g., metal objects, imaging device, viewpoints, lighting) significantly affects the specular reflection characteristics of individual metal object types. In this paper, we address this issue by introducing a tailor-made VDD framework that can be automatically adapted to a new domain. Specifically, we formulate this adaptation task as the problem of network architecture search (NAS) on a deep object-detection network, in which the network architecture is searched via reinforcement learning. We demonstrate the effectiveness of the proposed framework using the VDD-HPPPs task as a factory case study. Experimental results show that the proposed method achieved higher burr detection accuracy compared with the baseline method for data with different training/test domains for the non-convex HPPPs, which are particularly affected by domain shifts.

Koji Takeda, Kanji Tanaka

In visual robot self-localization, graph-based scene representation and matching have recently attracted research interest as robust and discriminative methods for selflocalization. Although effective, their computational and storage costs do not scale well to large-size environments. To alleviate this problem, we formulate self-localization as a graph classification problem and attempt to use the graph convolutional neural network (GCN) as a graph classification engine. A straightforward approach is to use visual feature descriptors that are employed by state-of-the-art self-localization systems, directly as graph node features. However, their superior performance in the original self-localization system may not necessarily be replicated in GCN-based self-localization. To address this issue, we introduce a novel teacher-to-student knowledge-transfer scheme based on rank matching, in which the reciprocal-rank vector output by an off-the-shelf state-of-the-art teacher self-localization model is used as the dark knowledge to transfer. Experiments indicate that the proposed graph-convolutional self-localization network can significantly outperform state-of-the-art self-localization systems, as well as the teacher classifier.

Tanaka Kanji, Yamaguchi Kousuke, Sugimoto Takuma

Loop-closure detection (LCD) in large non-stationary environments remains an important challenge in robotic visual simultaneous localization and mapping (vSLAM). To reduce computational and perceptual complexity, it is helpful if a vSLAM system has the ability to perform image change detection (ICD). Unlike previous applications of ICD, time-critical vSLAM applications cannot assume an offline background modeling stage, or rely on maintenance-intensive background models. To address this issue, we introduce a novel maintenance-free ICD framework that requires no background modeling. Specifically, we demonstrate that LCD can be reused as the main process for ICD with minimal extra cost. Based on these concepts, we develop a novel vSLAM component that enables simultaneous LCD and ICD. ICD experiments based on challenging cross-season LCD scenarios validate the efficacy of the proposed method.

Murase Tomoya, Tanaka Kanji

This paper addresses the problem of change detection from a novel perspective of long-term map learning. We are particularly interested in designing an approach that can scale to large maps and that can function under global uncertainty in the viewpoint (i.e., GPS-denied situations). Our approach, which utilizes a compact bag-of-words (BoW) scene model, makes several contributions to the problem: 1) Two kinds of prior information are extracted from the view sequence map and used for change detection. Further, we propose a novel type of prior, called motion prior, to predict the relative motions of stationary objects and anomaly ego-motion detection. The proposed prior is also useful for distinguishing stationary from non-stationary objects. 2) A small set of good reference images (e.g., 10) are efficiently retrieved from the view sequence map by employing the recently developed Bag-of-Local-Convolutional-Features (BoLCF) scene model. 3) Change detection is reformulated as a scene retrieval over these reference images to find changed objects using a novel spatial Bag-of-Words (SBoW) scene model. Evaluations conducted of individual techniques and also their combinations on a challenging dataset of highly dynamic scenes in the publicly available Malaga dataset verify their efficacy.

Fei Xiaoxiao, Tanaka Kanji, Inamoto Kouya

In this study, we explore the use of deep convolutional neural networks (DCNNs) in visual place classification for robotic mapping and localization. An open question is how to partition the robot's workspace into places to maximize the performance (e.g., accuracy, precision, recall) of potential DCNN classifiers. This is a chicken and egg problem: If we had a well-trained DCNN classifier, it is rather easy to partition the robot's workspace into places, but the training of a DCNN classifier requires a set of pre-defined place classes. In this study, we address this problem and present several strategies for unsupervised discovery of place classes ("time cue," "location cue," "time-appearance cue," and "location-appearance cue"). We also evaluate the efficacy of the proposed methods using the publicly available University of Michigan North Campus Long-Term (NCLT) Dataset.

Kanji Tanaka

In the classical context of robotic mapping and localization, map matching is typically defined as the task of finding a rigid transformation (i.e., 3DOF rotation/translation on the 2D moving plane) that aligns the query and reference maps built by mobile robots. This definition is valid in loop-rich trajectories that enable a mapper robot to close many loops, for which precise maps can be assumed. The same cannot be said about the newly emerging autonomous navigation and driving systems, which typically operate in loop-less trajectories that have no large loop (e.g., straight paths). In this paper, we propose a solution that overcomes this limitation by merging the two maps. Our study is motivated by the observation that even when there is no large loop in either the query or reference map, many loops can often be obtained in the merged map. We add two new aspects to map matching: (1) image retrieval with discriminative deep convolutional neural network (DCNN) features, which efficiently generates a small number of good initial alignment hypotheses; and (2) map merge, which jointly deforms the two maps to minimize differences in shape between them. To realize practical computation time, we also present a preemption scheme that avoids excessive evaluation of useless map-matching hypotheses. To verify our approach experimentally, we created a novel collection of uncertain loop-less maps by utilizing the recently published North Campus Long-Term (NCLT) dataset and its ground-truth GPS data. The results obtained using these map collections confirm that our approach improves on previous map-matching approaches.

Kanji Tanaka

Loop closure detection, which is the task of identifying locations revisited by a robot in a sequence of odometry and perceptual observations, is typically formulated as a visual place recognition (VPR) task. However, even state-of-the-art VPR techniques generate a considerable number of false positives as a result of confusing visual features and perceptual aliasing. In this paper, we propose a robust incremental framework for loop closure detection, termed incremental loop closure verification. Our approach reformulates the problem of loop closure detection as an instance of a multi-model hypothesize-and-verify framework, in which multiple loop closure hypotheses are generated and verified in terms of the consistency between loop closure hypotheses and VPR constraints at multiple viewpoints along the robot's trajectory. Furthermore, we consider the general incremental setting of loop closure detection, in which the system must update both the set of VPR constraints and that of loop closure hypotheses when new constraints or hypotheses arrive during robot navigation. Experimental results using a stereo SLAM system and DCNN features and visual odometry validate effectiveness of the proposed approach.

Tomoya Murase, Kanji Tanaka

Change detection, or anomaly detection, from street-view images acquired by an autonomous robot at multiple different times, is a major problem in robotic mapping and autonomous driving. Formulation as an image comparison task, which operates on a given pair of query and reference images is common to many existing approaches to this problem. Unfortunately, providing relevant reference images is not straightforward. In this paper, we propose a novel formulation for change detection, termed compressive change retrieval, which can operate on a query image and similar reference images retrieved from the web. Compared to previous formulations, there are two sources of difficulty. First, the retrieved reference images may frequently contain non-relevant reference images, because even state-of-the-art place-recognition techniques suffer from retrieval noise. Second, image comparison needs to be conducted in a compressed domain to minimize the storage cost of large collections of street-view images. To address the above issues, we also present a practical change detection algorithm that uses compressed bag-of-words (BoW) image representation as a scalable solution. The results of experiments conducted on a practical change detection task, "moving object detection (MOD)," using the publicly available Malaga dataset validate the effectiveness of the proposed approach.

Taisho Tsukamoto, Kanji Tanaka

In this study, we aim to solve the single-view robot self-localization problem by using visual experience across domains. Although the bag-of-words method constitutes a popular approach to single-view localization, it fails badly when it's visual vocabulary is learned and tested in different domains. Further, we are interested in using a cross-domain setting, in which the visual vocabulary is learned in different seasons and routes from the input query/database scenes. Our strategy is to mine a cross-domain visual experience, a library of raw visual images collected in different domains, to discover the relevant visual patterns that effectively explain the input scene, and use them for scene retrieval. In particular, we show that the appearance and the pose of the mined visual patterns of a query scene can be efficiently and discriminatively matched against those of the database scenes by employing image-to-class distance and spatial pyramid matching. Experimental results obtained using a novel cross-domain dataset show that our system achieves promising results despite our visual vocabulary being learned and tested in different domains.

Enfu Liu, Kanji Tanaka

Map retrieval, the problem of similarity search over a large collection of 2D pointset maps previously built by mobile robots, is crucial for autonomous navigation in indoor and outdoor environments. Bag-of-words (BoW) methods constitute a popular approach to map retrieval; however, these methods have extremely limited descriptive ability because they ignore the spatial layout information of the local features. The main contribution of this paper is an extension of the bag-of-words map retrieval method to enable the use of spatial information from local features. Our strategy is to explicitly model a unique viewpoint of an input local map; the pose of the local feature is defined with respect to this unique viewpoint, and can be viewed as an additional invariant feature for discriminative map retrieval. Specifically, we wish to determine a unique viewpoint that is invariant to moving objects, clutter, occlusions, and actual viewpoints. Hence, we perform scene parsing to analyze the scene structure, and consider the "center" of the scene structure to be the unique viewpoint. Our scene parsing is based on a Manhattan world grammar that imposes a quasi-Manhattan world constraint to enable the robust detection of a scene structure that is invariant to clutter and moving objects. Experimental results using the publicly available radish dataset validate the efficacy of the proposed approach.

Kanji Tanaka, Eiji Kondo

Vehicle relocation is the problem in which a mobile robot has to estimate the self-position with respect to an a priori map of landmarks using the perception and the motion measurements without using any knowledge of the initial self-position. Recently, RANdom SAmple Consensus (RANSAC), a robust multi-hypothesis estimator, has been successfully applied to offline relocation in static environments. On the other hand, online relocation in dynamic environments is still a difficult problem, for available computation time is always limited, and for measurement include many outliers. To realize real time algorithm for such an online process, we have developed an incremental version of RANSAC algorithm by extending an efficient preemption RANSAC scheme. This novel scheme named incremental RANSAC is able to find inlier hypotheses of self-positions out of large number of outlier hypotheses contaminated by outlier measurements.