Pith. sign in

REVIEW 2 major objections 4 minor 194 references

Mobile robots gain the contextual awareness needed for safe, targeted collaboration by combining continual person re-identification with multi-level geometric and semantic mapping.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-14 12:13 UTC pith:OGKU2MV6

load-bearing objection Solid engineering thesis that packages useful Re-ID and indoor LiDAR improvements; the long-term forgetting claim is the softest link but does not sink the rest. the 2 major comments →

arxiv 2607.10372 v1 pith:OGKU2MV6 submitted 2026-07-11 cs.RO cs.CV

Robotic Contextual Awareness for Human-Robot Collaboration and Environmental Understanding

Federico Rollo This is my paper
classification cs.RO cs.CV
keywords Context AwarenessEnvironmental UnderstandingRe-IdentificationSLAMLoop Closure DetectionSemantic MappingHuman-Robot CollaborationLiDAR
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This thesis argues that robots leaving controlled factories for dynamic human spaces fail without true contextual awareness: the ability to know both who a specific collaborator is and what the surrounding geometry and objects mean. It supplies that awareness through two complementary lines of work. On the human side, a visual re-identification pipeline (FollowMe, then CARPE-ID, then an unsupervised twin-network continual learner) lets a mobile robot lock onto one person, adapt to clothing and appearance changes, and ignore everyone else, while hand-gesture commands give simple interaction. On the environment side, a submap-based LiDAR SLAM system (LEO-SLAM) with multi-level scan matching, a ground-aware intensity filter that removes glass reflections without discarding floor points, and a Gaussian Scan Context loop-closure detector produce accurate maps; a multi-modal RGB-D-plus-LiDAR semantic layer then places named objects into those maps so the robot can navigate to them and perform simple fetch tasks. Together the methods turn raw sensor streams into a persistent, identity-aware and object-aware model of the scene, which the author claims is the missing prerequisite for safer coexistence and more natural collaboration.

Core claim

Comprehensive robotic contextual awareness arises from the joint solution of two data-association problems—persistent person re-identification under appearance change and robust geometric-plus-semantic place recognition—and that both can be solved online on mobile platforms by combining statistical model adaptation, multi-level submap matching, ground-aware intensity filtering, Gaussian Scan Context, and distance-adaptive camera-LiDAR fusion.

What carries the argument

The twin-network unsupervised continual learner that trains a parallel feature extractor on a smart image pool (recent target views mixed with statistically distant past appearances) while the live tracker continues uninterrupted, together with the multi-level submap alignment (scan-to-scan, scan-to-submaps, submap-to-submaps) and Gaussian Scan Context that supply the geometric backbone.

Load-bearing premise

The online twin-network training fed by the smart image pool is assumed to keep enough memory of earlier target appearances to avoid catastrophic forgetting while still running in real time without overflowing typical mobile-robot GPUs.

What would settle it

A long-duration person-following trial in which the target repeatedly changes clothing and reappears after multi-minute occlusions, measuring whether re-identification success rate remains high and whether tracking latency stays within real-time bounds on the same embedded hardware used in the thesis experiments.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. This doctoral thesis addresses robotic contextual awareness for mobile robots in human-centric settings through two complementary pillars: (1) visual person re-identification and tracking for targeted Human-Robot Collaboration (FollowMe baseline, CARPE-ID continual adaptation via DEMA-updated statistical models, and a twin-network unsupervised continual-learning extension with smart image-pool selection and Soft-Triplet loss to mitigate catastrophic forgetting), and (2) geometric and semantic environmental perception (LEO-SLAM multi-level submap scan matching with submap-based SC++, ground-aware intensity filtering for reflections, Gaussian Scan Context for robust loop closure, and multi-modal RGB-D/LiDAR Artifacts Mapping with distance-dependent fusion and data-association buffers). Methods are detailed in Chapter 4 with equations for statistical distances (Eq. 4.1), DEMA updates (Eqs. 4.5–4.8), multi-level GICP alignments, GSC matrix construction (Eq. 4.28), and camera-LiDAR fusion weights (Eqs. 4.36–4.37). Experiments in Chapter 5 report quantitative metrics (classification accuracy, tracking times, ATE/RPE on VBR, precision-recall for GSC, object detection rates) against SoTA baselines on public sequences and real-robot datasets, plus qualitative HRC and loco-manipulation demos.

Significance. If the integrated claims hold, the work supplies a practical, modular suite of perception modules that advance mobile robots from purely geometric navigation toward identity-aware, semantically enriched operation in unstructured human environments. Strengths include real-robot validation (quadruped platforms, custom indoor datasets with glass reflections), ablations (Table 5.2 on statistical-model updates and early training), public-benchmark comparisons (VBR ATE/RPE in Table 5.3, GSC recall under perfect precision in Table 5.5), and explicit engineering for online constraints (parallel twin network, submap keyframes, intensity ground awareness). The multi-modal fusion and GSC statistical extension are particularly transferable. The thesis format naturally aggregates prior conference results into a coherent narrative of contextual awareness, which is valuable for the robotics community even if individual modules are incremental.

major comments (2)
  1. Section 4.1.3 and Section 5.1.3 (Table 5.2, Fig. 5.5, saliency maps in Fig. 5.6): The central HRC claim that the twin-network continual learner plus smart image-pool selection yields a 'highly personalized and robust' long-term Re-ID system rests on the assumption that catastrophic forgetting of earlier appearances is prevented while real-time tracking continues uninterrupted. The reported experiments measure only short-horizon mean tracking times and Re-ID success relative to MOT failures, plus qualitative Grad-CAM maps; they lack multi-session sequences with deliberate multi-outfit changes, quantitative retention curves for early appearances after successive weight swaps, and measured peak GPU memory on representative mobile-robot hardware. Without these, the long-term robustness asserted in the abstract and Section 1.3.1 remains under-supported for the multi-day HRC scenarios claimed.
  2. Section 1.2–1.3 and Chapter 6: The thesis repeatedly asserts that the two pillars 'synergistically' produce comprehensive contextual awareness enabling safer coexistence and more effective collaboration. Yet the experimental chapters evaluate the Re-ID, LEO-SLAM/GSC, and Artifacts Mapping pipelines largely in isolation (separate tables and figures); the only joint demonstration is a high-level 'bring-me' loco-manipulation sketch (Fig. 4.17) that does not quantify interaction between person identity and semantic map. A load-bearing integrated experiment or explicit cross-module ablation is needed to substantiate the synergistic claim that underpins the title and abstract.
minor comments (4)
  1. Multiple typographical and formatting issues appear throughout: 'Accademic Advisor', 'A w areness', 'two-dimentional', 'catastrofic forgetting', 'simultaneous' misspellings, and inconsistent capitalization of acronyms (e.g., SoTA vs SOTA). A thorough proof-reading pass is required.
  2. Figures 4.10, 4.14 and several experimental plots are dense; axis labels and legend fonts are sometimes too small for print readability. Consider enlarging key panels or providing higher-resolution versions.
  3. Free parameters (λ_d construction, DEMA Δ_f/Δ_λd, GSC α and Huber δ, fusion breakpoints min_c/acc_c/max_c, batch size N) are listed but sensitivity analyses are sparse outside the CARPE-ID ablation. A short appendix table summarizing default values and observed sensitivity would aid reproducibility.
  4. Related-work coverage of open-world semantic mapping (Section 2.3) is up-to-date but the thesis itself remains closed-set; a brief forward-looking paragraph on how the Artifacts Mapping pipeline could incorporate CLIP-style embeddings would strengthen the discussion.

Circularity Check

0 steps flagged

No circularity: empirical systems with free parameters and external-benchmark evaluations; self-citations are normal thesis reuse of the author's own conference experiments, not load-bearing derivations.

full rationale

The thesis proposes algorithmic pipelines (FollowMe/CARPE-ID + twin-network continual learning, LEO-SLAM multi-level submap matching, ground-aware intensity filter, Gaussian Scan Context, multi-modal Artifacts Mapping) and evaluates them on real-robot trajectories, VBR/KITTI sequences, precision-recall curves, tracking-time tables, and human-labeled Re-ID success. Thresholds (λ_d, damping factors, intensity bounds, Huber δ, fusion weights ξ) are free design choices set by calibration or heuristics; reported metrics (ATE/RPE, recall@100% precision, mean tracking time, detection accuracy) are measured against independent ground truth or MOT failures, not forced by construction from the same inputs. Self-citations ([111],[112],[110],[107],[109]) simply point to the conference versions of the same experiments; they do not import uniqueness theorems, smuggle ansätze, or redefine the claimed performance. No equation reduces a 'prediction' to a fitted quantity by identity, and no central claim rests solely on an unverified self-citation chain. The work is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

6 free parameters · 4 axioms · 4 invented entities

The work is engineering-heavy: most free parameters are classic robotics thresholds chosen by hand or by small validation sets; axioms are standard SLAM and computer-vision assumptions; invented entities are the named frameworks themselves, which are methods rather than physical postulates and therefore lack independent external evidence beyond the experiments in this thesis and its constituent papers.

free parameters (6)
  • Re-ID distance threshold λ_d = μ_d + 2σ_d and DEMA damping factors Δ_f, Δ_λd
    Chosen to balance false positives/negatives; updated online but initial values and N_max are hand-set.
  • Submap creation thresholds Δ_m (distance) and Δ_θ (rotation)
    Control keyframe density; selected for the target environments without automatic tuning.
  • Intensity bounds ψ*_min, ψ*_max and ground-height check
    Sensor- and environment-specific; critical for the ground-aware filter performance.
  • GSC variance weight α and Huber threshold δ
    Tunable hyper-parameters that trade robustness versus discrimination in the context matrix.
  • Camera-LiDAR fusion breakpoints min_c, acc_c, max_c and weight ξ
    Taken from vendor specs or set by hand; directly affect the piecewise centroid fusion.
  • Continual-learning batch size N, iteration count, Soft-Triplet margin α, early-training timeout
    Control training frequency and memory; affect both accuracy and real-time feasibility.
axioms (4)
  • domain assumption Markov assumption for robot motion and conditional independence of observations given pose and map (standard SLAM factorization).
    Invoked throughout Section 3.2 and the LEO-SLAM factor-graph construction.
  • domain assumption LiDAR intensity is a reliable proxy for material reflectance and incidence angle, allowing thresholding to remove reflections while ground points can be recovered by height or normal checks.
    Underpins the ground-aware filter (Section 4.2.2) and is only partially validated by the custom indoor glass-door dataset.
  • ad hoc to paper A statistical feature-space distance (normalized Euclidean) plus a continually updated mean/std model is sufficient to re-identify a person under moderate appearance change.
    Core of FollowMe and CARPE-ID; justified empirically but not derived from first principles.
  • ad hoc to paper Submap aggregation of successive LiDAR scans yields descriptors comparable to high-channel outdoor LiDAR for indoor loop closure.
    Stated in Section 4.2.1 as the motivation for submap-based SC++.
invented entities (4)
  • CARPE-ID framework (with DEMA adaptation and later twin-network continual learning) no independent evidence
    purpose: Personalized online person re-identification that adapts to appearance change and mitigates catastrophic forgetting.
    Named system introduced in the thesis; evaluated only within the author's experiments.
  • LEO-SLAM (multi-level submap alignment + submap SC++) no independent evidence
    purpose: Accurate 3D LiDAR mapping and localization with reduced drift.
    New pipeline; performance shown on VBR and custom data only.
  • Gaussian Scan Context (GSC) and Huber-weighted variant no independent evidence
    purpose: More robust place-recognition descriptor than max-height SC++.
    Statistical reformulation of an existing descriptor; no external re-implementation cited.
  • Artifacts Mapping multi-modal semantic pipeline no independent evidence
    purpose: Online object-of-interest mapping via distance-adaptive camera-LiDAR fusion.
    Custom fusion rule and manager; validated on simulation and real scenes by the author.

pith-pipeline@v1.1.0-grok45 · 51491 in / 3494 out tokens · 45293 ms · 2026-07-14T12:13:33.500751+00:00 · methodology

0 comments
read the original abstract

The transition of autonomous mobile robots from controlled industrial settings to dynamic, human-centric environments, such as manufacturing, logistics, and healthcare, has made their safe and autonomous operation a critical area of research. These sophisticated machines must be capable of perceiving, understanding, and interacting with their surroundings to navigate freely and perform complex tasks. A significant obstacle to achieving this is the lack of comprehensive contextual awareness, which requires a robot to recognize its spatial environment and identify the objects and actors within it. Without this perceptual knowledge, robots struggle to plan adaptive behaviors or engage in meaningful interaction with humans. This thesis presents novel solutions to this challenge by exploring two distinct but complementary research directions. The first direction involves human re-identification and tracking to improve Human-Robot Collaboration. Our developed approach enables a mobile robot to recognize a specific person, facilitating targeted collaboration while ignoring other individuals. The second direction focuses on enhancing the robot's overall perceptual capabilities to understand its environment geometrically and semantically. Geometric information is vital for motion planning and collision avoidance, while semantic knowledge provides the robot with a richer understanding for more advanced interaction. Both solutions are driven by the improvement of the semantical understanding of robots that enhance their knowledge of their surroundings, allowing a smoother and more natural interaction between robots, humans, and the environment. The contributions of this work in human re-identification and environmental understanding represent a significant step toward a future where robots are more contextually aware, enabling safer coexistence and more effective collaboration.

Figures

Figures reproduced from arXiv: 2607.10372 by Federico Rollo.

Figure 2.1
Figure 2.1. Figure 2.1: Comparison of LiDAR-perceived intensities across diverse materials. The intensity ranges from the lower values of shiny metallic surfaces, represented in light blue, through the medium values of opaque metallic surfaces in violet, to the higher values of the wooden panel in pink. Several studies have utilized intensity data to improve SLAM components. Wang et al. [137] proposed the Intensity Scan Context… view at source ↗
Figure 2.2
Figure 2.2. Figure 2.2: Examples of the filtering effect are shown in two scenarios that show numer [PITH_FULL_IMAGE:figures/full_fig_p044_2_2.png] view at source ↗
Figure 2.3
Figure 2.3. Figure 2.3: The incidence angle θ between the LiDAR beam and the reflective surface significantly impacts the returned energy level, often reducing the point intensity. Points affected in this way would typically be removed by standard threshold filtering methods. Intensity-based filtering has previously been investigated in autonomous driving ap￾plications to remove noise caused by weather phenomena like snow or ra… view at source ↗
Figure 4.1
Figure 4.1. Figure 4.1: FollowMe perception pipeline: (a) RGB and Depth images acquisition; (b) person detection through Yolact++; (c) person Re-ID using a deep neural network and human features distance; (d) person localization using the point cloud and re-identified person mask; (e) gesture detection to send commands to the robot. Reprinted, with permission, from [111], © 2023 IEEE. People detection To detect individuals in t… view at source ↗
Figure 4.2
Figure 4.2. Figure 4.2: In green, the plot of the feature distances for a calibrated person used to [PITH_FULL_IMAGE:figures/full_fig_p066_4_2.png] view at source ↗
Figure 4.3
Figure 4.3. Figure 4.3: Overview of relevant transformation frames and representation of the safety [PITH_FULL_IMAGE:figures/full_fig_p068_4_3.png] view at source ↗
Figure 4.4
Figure 4.4. Figure 4.4: Classes considered to train the SVM for hand gesture detection. Mediapipe framework [157] is used to extract 3D key points (red points) relative to the centre of the hand. Reprinted, with permission, from [111], © 2023 IEEE. A radial basis function kernel and a one-versus-one decision function were employed to train the multi-class SVM classifier. The classification output is filtered (i.e., ξ equal cons… view at source ↗
Figure 4.5
Figure 4.5. Figure 4.5: FollowMe FSM diagram. The transitions between the states are highlighted with Greek letters and boolean operators and (&), not (!). The transitions’ meaning is: (α) the target is re-identified in the camera FOV, (β) the last target position is inside the safety circle, and (γ) the robot receives a command by the hand gestures module. Reprinted, with permission, from [111], © 2023 IEEE. The safety distanc… view at source ↗
Figure 4.6
Figure 4.6. Figure 4.6: The pipeline framework begins with an image input and ends with a re [PITH_FULL_IMAGE:figures/full_fig_p071_4_6.png] view at source ↗
Figure 4.7
Figure 4.7. Figure 4.7: This figure illustrates a comparison between [PITH_FULL_IMAGE:figures/full_fig_p073_4_7.png] view at source ↗
Figure 4.8
Figure 4.8. Figure 4.8: This work pipeline, integrated with the CARPE-ID framework, outlines the algorithmic flow from the input image to the Re-ID output. The arrows symbolize the data exchange among the various components within the framework. Reprinted, with permission, from [110], © 2025 IEEE. To address this issue, an autonomous continual learning module is presented, aimed at fine-tuning the feature extraction network to … view at source ↗
Figure 4.9
Figure 4.9. Figure 4.9: On the left, the pool collection of training images for adapting the twin [PITH_FULL_IMAGE:figures/full_fig_p078_4_9.png] view at source ↗
Figure 4.10
Figure 4.10. Figure 4.10: LEO-SLAM system architecture. The algorithm receives as input a point cloud P and optionally external odometry or IMU data. The point cloud is prepro￾cessed, and the output Pp is passed in parallel into the scan-to-scan (s2s) and scan-to￾submaps (s2S) alignment modules, as well as into a filtering module, which produces Pf , later used for building the final map. If the robot has travelled a distance dm… view at source ↗
Figure 4.11
Figure 4.11. Figure 4.11: Comparison of intensity filtering: on the left, using only intensity thresh [PITH_FULL_IMAGE:figures/full_fig_p085_4_11.png] view at source ↗
Figure 4.12
Figure 4.12. Figure 4.12: The diagonal and top views of a GSC, where each entry represents a multivariate normal distribution, are shown on the left and right, respectively. Each Gaussian distribution contributes to the computation of the GSC matrix. To compute this angle, the vector cosine similarity measure, θηi , is used, which is given by: θηi = ⟨ fηi , f z⟩ ∥ fηi∥2∥ fz∥2 (4.18) where ⟨·, ·⟩ denotes the dot product , and ∥x∥… view at source ↗
Figure 4.13
Figure 4.13. Figure 4.13: The plots compare SC++ (left) and GSC (right) during a loop closure in the KITTI sequence 00. The GSC matrix, displayed at the lower right, is derived from the Gaussian distributions illustrated in [PITH_FULL_IMAGE:figures/full_fig_p089_4_13.png] view at source ↗
Figure 4.14
Figure 4.14. Figure 4.14: This figure represents the whole Artifacts Mapping pipeline. The top block [PITH_FULL_IMAGE:figures/full_fig_p095_4_14.png] view at source ↗
Figure 4.15
Figure 4.15. Figure 4.15: An example of the contribution weights of camera and [PITH_FULL_IMAGE:figures/full_fig_p098_4_15.png] view at source ↗
Figure 4.16
Figure 4.16. Figure 4.16: An example of the framework during an experiment. On the left is the [PITH_FULL_IMAGE:figures/full_fig_p100_4_16.png] view at source ↗
Figure 4.17
Figure 4.17. Figure 4.17: Autonomous actions held during the execution of the application. Image [PITH_FULL_IMAGE:figures/full_fig_p101_4_17.png] view at source ↗
Figure 5.1
Figure 5.1. Figure 5.1: Comparison between panoptic and instance segmentation inferences. [PITH_FULL_IMAGE:figures/full_fig_p106_5_1.png] view at source ↗
Figure 5.2
Figure 5.2. Figure 5.2: Results of the FollowMe whole experiment: each person has to follow an [PITH_FULL_IMAGE:figures/full_fig_p108_5_2.png] view at source ↗
Figure 5.3
Figure 5.3. Figure 5.3: Individual experiments statistics. Reprinted, with permission, from [ [PITH_FULL_IMAGE:figures/full_fig_p110_5_3.png] view at source ↗
Figure 5.4
Figure 5.4. Figure 5.4: This is an example of a person-following experiment. The person being [PITH_FULL_IMAGE:figures/full_fig_p111_5_4.png] view at source ↗
Figure 5.5
Figure 5.5. Figure 5.5: Comparison between the experimental results of [PITH_FULL_IMAGE:figures/full_fig_p115_5_5.png] view at source ↗
Figure 5.6
Figure 5.6. Figure 5.6: Saliency maps comparison obtained with Grad-Cam [ [PITH_FULL_IMAGE:figures/full_fig_p116_5_6.png] view at source ↗
Figure 5.7
Figure 5.7. Figure 5.7: Comparison between the trajectories generated by LEO [PITH_FULL_IMAGE:figures/full_fig_p119_5_7.png] view at source ↗
Figure 5.8
Figure 5.8. Figure 5.8: Final maps obtained using LEO-SLAM from the four analyzed VBR ses￾sions. Reprinted, with permission, from [107], © 2025 IEEE. AT E = vuut 1 N X N i=1 ||pˆi − pi ||2 , (5.1) RP E = vuut 1 N X N i=1 ||(pˆi+1 − pˆi ) − (pi+1 − pi )||2 , (5.2) where pˆi denotes the estimated trajectory position at time i, pi is the corresponding ground-truth position at time i, and N is the total number of estimated trajecto… view at source ↗
Figure 5.9
Figure 5.9. Figure 5.9: Precision–Recall curves obtained from the experimental results are pre [PITH_FULL_IMAGE:figures/full_fig_p126_5_9.png] view at source ↗
Figure 5.10
Figure 5.10. Figure 5.10: Percentage of the correctly mapped and labelled objects concerning the [PITH_FULL_IMAGE:figures/full_fig_p132_5_10.png] view at source ↗
Figure 5.11
Figure 5.11. Figure 5.11: Distribution of correctly and wrongly detected artifacts among the total [PITH_FULL_IMAGE:figures/full_fig_p133_5_11.png] view at source ↗
Figure 5.12
Figure 5.12. Figure 5.12: Qualitative comparison between RGB-D camera (left image) and LiDAR (right image) point cloud detections at an approximate distance of 10 m from the wall. At the bottom center is a representation of the scene taken with the robot camera at that instant. At large distances, camera data are noisier and less accurate than the LiDAR one. Still, at short distances, cameras provide a denser, more accurate poin… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

194 extracted references · 15 linked inside Pith

  1. [1]

    IEEE trans on pattern analysis and machine intel , year=

    Yolact++: Better real-time instance segmentation , author=. IEEE trans on pattern analysis and machine intel , year=

  2. [2]

    arXiv preprint arXiv:2001.01526 , year=

    Mutual mean-teaching: Pseudo label refinery for unsupervised domain adaptation on person re-identification , author=. arXiv preprint arXiv:2001.01526 , year=

    Pith/arXiv arXiv 2001

  3. [3]

    arXiv preprint arXiv:2006.10214 , year=

    Mediapipe hands: On-device real-time hand tracking , author=. arXiv preprint arXiv:2006.10214 , year=

    Pith/arXiv arXiv 2006

  4. [4]

    IEEE/RSJ Int Conf on Intel Robots and sys , year=

    Followme: Person following and gesture recognition with a quadrocopter , author=. IEEE/RSJ Int Conf on Intel Robots and sys , year=

  5. [5]

    ICT Systems and Sustainability: Proceedings of ICT4SD 2020, Volume 1 , pages=

    Follow me: A human following robot using wi-fi received signal strength indicator , author=. ICT Systems and Sustainability: Proceedings of ICT4SD 2020, Volume 1 , pages=. 2020 , publisher=

    2020

  6. [6]

    IEEE/SICE Int Symp on Sys Integration (SII) , year=

    Human characterization by a following robot using a depth sensor , author=. IEEE/SICE Int Symp on Sys Integration (SII) , year=

  7. [7]

    Int Conf on Control, Automation and Sys , year=

    Human-following robot using infrared camera , author=. Int Conf on Control, Automation and Sys , year=

  8. [8]

    Int Conf on Intel Autonomous Sys , year=

    Tracking control of human-following robot with sonar sensors , author=. Int Conf on Intel Autonomous Sys , year=

  9. [9]

    Int Conf on Orange Technologies (ICOT) , year=

    Design and implementation of human following for separable omnidirectional mobile system of smart home robot , author=. Int Conf on Orange Technologies (ICOT) , year=

  10. [10]

    IEEE Trans on Consumer Electronics , volume=

    Tracking autonomous entities using RFID technology , author=. IEEE Trans on Consumer Electronics , volume=

  11. [11]

    South East Asian Tech University Consortium (SEATUC) , volume=

    Simulation of a human following robot with object avoidance function , author=. South East Asian Tech University Consortium (SEATUC) , volume=

  12. [12]

    IEEE computer society conf on computer vision and pattern recognition (CVPR'05) , volume=

    Histograms of oriented gradients for human detection , author=. IEEE computer society conf on computer vision and pattern recognition (CVPR'05) , volume=

  13. [13]

    CHI'12 Extended Abstracts on Human Factors in Computing Sys , year=

    Joggobot: a flying robot as jogging companion , author=. CHI'12 Extended Abstracts on Human Factors in Computing Sys , year=

  14. [14]

    IEEE/RSJ Int Conf on Intel Robots and Sys (IROS) , year=

    User recognition for guiding and following people with a mobile robot in a clinical environment , author=. IEEE/RSJ Int Conf on Intel Robots and Sys (IROS) , year=

  15. [15]

    Int Conf on Advances in Computing, Communications and Informatics (ICACCI) , year=

    Follow me robot using bluetooth-based position estimation , author=. Int Conf on Advances in Computing, Communications and Informatics (ICACCI) , year=

  16. [16]

    Robot World Cup , year=

    Follow Me: real-time in the wild person tracking application for autonomous robotics , author=. Robot World Cup , year=

  17. [17]

    Academic Research International , volume=

    Follow me robot using infrared beacons , author=. Academic Research International , volume=. 2013 , publisher=

    2013

  18. [18]

    2008 , publisher=

    Person following robot with vision-based and sensor fusion tracking algorithm , author=. 2008 , publisher=

    2008

  19. [19]

    , author=

    Vision based person tracking with a mobile robot. , author=. BMVC , year=

  20. [20]

    Gita and Kilo, Piaggio Fast Forward

  21. [21]

    Robotic Cart Thouzer, Doog inc

  22. [22]

    Proc of the IEEE int conf on computer vision , year=

    Mask r-cnn , author=. Proc of the IEEE int conf on computer vision , year=

  23. [23]

    IEEE Trans on Pattern Analysis and Machine Intelligence , year=

    Deep learning for person re-identification: A survey and outlook , author=. IEEE Trans on Pattern Analysis and Machine Intelligence , year=

  24. [24]

    IEEE Int Conf on Power Electronics, Intel Control and Energy Sys (ICPEICES) , year=

    A review: Study of various techniques of Hand gesture recognition , author=. IEEE Int Conf on Power Electronics, Intel Control and Energy Sys (ICPEICES) , year=

  25. [25]

    Proc of the European Conf on Computer Vision (ECCV) , year=

    Two at once: Enhancing learning and generalization capacities via ibn-net , author=. Proc of the European Conf on Computer Vision (ECCV) , year=

  26. [26]

    Proc of the IEEE conf on computer vision and pattern recognition (CVPR) , year=

    Person transfer gan to bridge domain gap for person re-identification , author=. Proc of the IEEE conf on computer vision and pattern recognition (CVPR) , year=

  27. [27]

    IEEE/RSJ Int Conf on Intel Robots and Sys (IROS) , year=

    A visuo-haptic guidance interface for mobile collaborative robotic assistant (MOCA) , author=. IEEE/RSJ Int Conf on Intel Robots and Sys (IROS) , year=

  28. [28]

    Proc of the IEEE conf on computer vision and pattern recognition , year=

    Realtime multi-person 2d pose estimation using part affinity fields , author=. Proc of the IEEE conf on computer vision and pattern recognition , year=

  29. [29]

    ICRA workshop on open source software , year=

    ROS: an open-source Robot Operating Sys , author=. ICRA workshop on open source software , year=

  30. [30]

    IEEE Access , volume=

    LMOT: Efficient Light-Weight Detection and Tracking in Crowds , author=. IEEE Access , volume=. 2022 , publisher=

    2022

  31. [31]

    Proc of the IEEE/CVF conf on computer vision and pattern recognition , year=

    Improving multiple pedestrian tracking by track management and occlusion handling , author=. Proc of the IEEE/CVF conf on computer vision and pattern recognition , year=

  32. [32]

    European conf on computer vision , year=

    Microsoft coco: Common objects in context , author=. European conf on computer vision , year=

  33. [33]

    Proc of the IEEE/CVF int conf on computer vision , year=

    Yolact: Real-time instance segmentation , author=. Proc of the IEEE/CVF int conf on computer vision , year=

  34. [34]

    IEEE Int Conf on Robotics and Automation (ICRA) , year=

    Yolactedge: Real-time instance segmentation on the edge , author=. IEEE Int Conf on Robotics and Automation (ICRA) , year=

  35. [35]

    Proc of the IEEE conf on computer vision and pattern recognition , year=

    You only look once: Unified, real-time object detection , author=. Proc of the IEEE conf on computer vision and pattern recognition , year=

  36. [36]

    Proc of the IEEE conf on computer vision and pattern recognition , year=

    Fully convolutional networks for semantic segmentation , author=. Proc of the IEEE conf on computer vision and pattern recognition , year=

  37. [37]

    Chinese Control Conf (CCC) , year=

    On adaptive monte carlo localization algorithm for the mobile robot based on ROS , author=. Chinese Control Conf (CCC) , year=

  38. [38]

    IEEE Int Symp on Wearable Computers, 2003

    Personal position measurement using dead reckoning , author=. IEEE Int Symp on Wearable Computers, 2003. Proc. , year=

    2003

  39. [39]

    IEEE int symp on robotics and manufacturing automation (ROMA) , year=

    A quantitative study of tuning ROS gmapping parameters and their effect on performing indoor 2D SLAM , author=. IEEE int symp on robotics and manufacturing automation (ROMA) , year=

  40. [40]

    2023 IEEE International Conference on Advanced Robotics and Its Social Impacts (ARSO) , pages=

    FollowMe: a Robust Person Following Framework Based on Visual Re-Identification and Gestures , author=. 2023 IEEE International Conference on Advanced Robotics and Its Social Impacts (ARSO) , pages=. 2023 , organization=

    2023

  41. [41]

    2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , year=

    Personalized Re-identification through Unsupervised Continual Learning and Parallel Training , author=. 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , year=

    2025

  42. [42]

    arXiv preprint arXiv:2201.13066 , year=

    Single Object Tracking: A Survey of Methods, Datasets, and Evaluation Metrics , author=. arXiv preprint arXiv:2201.13066 , year=

    Pith/arXiv arXiv

  43. [43]

    Neurocomputing , volume=

    Recent advances of single-object tracking methods: A brief survey , author=. Neurocomputing , volume=. 2021 , publisher=

    2021

  44. [44]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Visual object tracking with discriminative filters and siamese networks: a survey and outlook , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

  45. [45]

    Computer Vision and Image Understanding , volume=

    Visual object tracking: A survey , author=. Computer Vision and Image Understanding , volume=. 2022 , publisher=

    2022

  46. [46]

    IEEE Transactions on Multimedia , year=

    Strongsort: Make deepsort great again , author=. IEEE Transactions on Multimedia , year=

  47. [47]

    IEEE international conference on image processing (ICIP) , pages=

    Simple online and realtime tracking with a deep association metric , author=. IEEE international conference on image processing (ICIP) , pages=

  48. [48]

    European Conference on Computer Vision , year=

    Large scale Real-world Multi Person Tracking , author=. European Conference on Computer Vision , year=

  49. [49]

    IEEE/CVF Winter Conference on Applications of Computer Vision , pages=

    Mmptrack: Large-scale densely annotated multi-camera multiple people tracking benchmark , author=. IEEE/CVF Winter Conference on Applications of Computer Vision , pages=

  50. [50]

    IEEE International Conference on Robotics and Automation (ICRA) , year=

    Robot Person Following Under Partial Occlusion , author=. IEEE International Conference on Robotics and Automation (ICRA) , year=

  51. [51]

    CAAI Transactions on Intelligence Technology , year=

    Online RGB-D person re-identification based on metric model update , author=. CAAI Transactions on Intelligence Technology , year=

  52. [52]

    Pattern Recognition Letters , year=

    A real-time and unsupervised face re-identification system for human-robot interaction , author=. Pattern Recognition Letters , year=

  53. [53]

    Artificial intelligence , year=

    Multiple object tracking: A literature review , author=. Artificial intelligence , year=

  54. [54]

    Pattern Recognition , year=

    People re-identification using skeleton standard posture and color descriptors from RGB-D data , author=. Pattern Recognition , year=

  55. [55]

    Robotics and Autonomous Systems , year=

    Monocular person tracking and identification with on-line deep feature selection for person following robots , author=. Robotics and Autonomous Systems , year=

  56. [56]

    Journal of Intelligent & Robotic Systems , year=

    Human Re-identification with a robot thermal camera using entropy-based sampling , author=. Journal of Intelligent & Robotic Systems , year=

  57. [57]

    International Symposium on Robotics , pages=

    May I be your personal coach? bringing together person tracking and visual re-identification on a mobile robot , author=. International Symposium on Robotics , pages=. 2016 , organization=

    2016

  58. [58]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Alphapose: Whole-body regional multi-person pose estimation and tracking in real-time , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

  59. [59]

    Proceedings of the national academy of sciences , volume=

    Overcoming catastrophic forgetting in neural networks , author=. Proceedings of the national academy of sciences , volume=. 2017 , publisher=

    2017

  60. [60]

    International conference on machine learning , pages=

    Continual learning through synaptic intelligence , author=. International conference on machine learning , pages=. 2017 , organization=

    2017

  61. [61]

    arXiv preprint arXiv:1606.04671 , year=

    Progressive neural networks , author=. arXiv preprint arXiv:1606.04671 , year=

    Pith/arXiv arXiv

  62. [62]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Learning continual compatible representation for re-indexing free lifelong person re-identification , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  63. [63]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Lifelong person re-identification via adaptive knowledge accumulation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  64. [64]

    Conference on lifelong learning agents , pages=

    Continual novelty detection , author=. Conference on lifelong learning agents , pages=. 2022 , organization=

    2022

  65. [65]

    Conference on lifelong learning agents , pages=

    Fixed design analysis of regularization-based continual learning , author=. Conference on lifelong learning agents , pages=. 2023 , organization=

    2023

  66. [66]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    A comprehensive survey of continual learning: Theory, method and application , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2024 , publisher=

    2024

  67. [67]

    Continuous Adaptation in Person Re-identification for Robotic Assistance , year=

    Rollo, Federico and Zunino, Andrea and Tsagarakis, Nikolaos and Hoffman, Enrico Mingo and Ajoudani, Arash , booktitle=. Continuous Adaptation in Person Re-identification for Robotic Assistance , year=

  68. [68]

    Facenet: A unified embedding for face recognition and clustering , author=. conf. on computer vision and pattern recognition (CVPR) , year=

  69. [69]

    Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification , author=. Int. Conf. on Learning Representations , year=

  70. [70]

    Neural networks , year=

    Continual lifelong learning with neural networks: A review , author=. Neural networks , year=

  71. [71]

    Neurocomputing , year=

    Online continual learning in image classification: An empirical survey , author=. Neurocomputing , year=

  72. [72]

    Information fusion , year=

    Continual learning for robotics: Definition, framework, learning strategies, opportunities and challenges , author=. Information fusion , year=

  73. [73]

    Connection Science , year=

    Catastrophic forgetting, rehearsal and pseudorehearsal , author=. Connection Science , year=

  74. [74]

    Cognitive Computation , year=

    A bio-inspired incremental learning architecture for applied perceptual problems , author=. Cognitive Computation , year=

  75. [75]

    icarl: Incremental classifier and representation learning , author=. Conf. on Computer Vision and Pattern Recognition (CVPR) , year=

  76. [76]

    European Conference on Computer Vision (ECCV) , year=

    GDumb: A Simple Approach that Questions Our Progress in Continual Learning , author=. European Conference on Computer Vision (ECCV) , year=

  77. [77]

    Task-free continual learning , author=. Conf. on Computer Vision and Pattern Recognition (CVPR) , year=

  78. [78]

    international conference on computer vision (ICCV) , year=

    Grad-cam: Visual explanations from deep networks via gradient-based localization , author=. international conference on computer vision (ICCV) , year=

  79. [79]

    International Conference on Computer Vision (ICCV) , year=

    Softtriple loss: Deep metric learning without triplet sampling , author=. International Conference on Computer Vision (ICCV) , year=

  80. [80]

    AAAI conf

    Learning incremental triplet margin for person re-identification , author=. AAAI conf. on artificial intelligence , year=

Showing first 80 references.