Pith. sign in

REVIEW 5 major objections 6 minor 35 references

IRisPath: Enhancing Costmap for Off-Road Navigation with Robust IR-RGB Fusion for Improved Day and Night Traversability

T0 review · 5 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper claims that fusing long-wave infrared and RGB imagery, plus ego-velocity, produces traversability costmaps for off-road robots that stay reliable across day and night, according to qualitative comparisons with single-modality…

desk verdict The dataset and calibration idea are the real assets; the central fusion claim is unsupported by the evidence as presented. read the letter →

arxiv 2412.03173 v2 pith:LNAF4SAN submitted 2024-12-04 cs.RO cs.CV

Saksham Sharma , Akshit Raizada , Suresh Sundaram This is my paper · ORCID
classification cs.ROcs.CV
keywords traversabilitycostmapoff-roadnavigationthermal-RGBfusionself-supervisedlearningextrinsiccalibrationday-nightdatasetLWIRimageryego-velocityconditioning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

IRisPath claims that fusing RGB images with long-wave infrared (thermal) images, and conditioning on the robot's own speed, yields off-road traversability costmaps that stay accurate across day and night, where RGB-only or infrared-only models degrade. The method is self-supervised: labels come from the robot's vertical vibrations measured by an onboard IMU, so no manual terrain annotation is needed. It also introduces a targetless calibration procedure that aligns thermal and RGB cameras through a LiDAR intensity image, along with a new day-night off-road dataset containing co-registered thermal, RGB, LiDAR, and IMU data. If correct, off-road robots could maintain reliable risk maps in darkness, fog, and dust without retraining.

What carries the argument

The load-bearing mechanism is the self-supervised pseudo-label, defined as $y = \mathrm{PSD}(\mathrm{acc}_z) / \sqrt{V_x^2 + V_y^2 + 10}$, where $\mathrm{PSD}(\mathrm{acc}_z)$ is the area under the power spectral density of the robot's vertical acceleration and $V_x, V_y$ are horizontal velocity components. This quantity converts IMU vibration data into a per-image-patch traversability cost, and training patches are sampled from directly beneath the robot so the IMU reading corresponds to the exact terrain in the patch. Around this label, the model stacks two ResNet-18 encoders (one per modality), Fourier-encoded velocity, and a fusion MLP; the companion targetless calibration chain computes the RGB-to-thermal transform via LiDAR intensity images and reprojection-error minimization.

What would settle it

A direct test would be to collect data on a smooth but low-traction surface—for instance, wet clay or crushed gravel with low rolling resistance—and compare the model's predicted costmap against a hand-labelled or physically measured difficulty map. If the model assigns low cost where the vehicle visibly slips, struggles, or deviates from its commanded path, the IMU-vibration pseudo-label fails to capture a central component of traversability.

Watch

Extended reading notes

Core claim

The central claim is that a fusion model that processes RGB and LWIR imagery through separate feature encoders, injects ego-velocity through Fourier features, and regresses a per-patch traversability cost, outperforms either modality alone in both daylight and night conditions. The paper's qualitative results show that during the day the fused costmap assigns lower costs to traversable paths than an RGB-only model and sharper boundaries than an infrared-only model; at night, the fused map recovers terrain features that neither single modality captures—distant boundaries from the thermal channel and close-range details like grass and roots from the RGB channel. The authors present this as evidence that complementary visual modalities make self-supervised costmap learning more robust to lighting and weather variation.

Load-bearing premise

The load-bearing premise is that the area under the power spectral density of the robot's vertical acceleration, normalized by horizontal velocity, faithfully measures how difficult a terrain patch is to traverse.

Editorial extensions

If this is right

  • Off-road robots can keep building usable costmaps through day-night transitions and in low-visibility conditions without retraining on the new lighting regime.
  • The open-source day-night dataset with thermal, RGB, LiDAR, and IMU data gives other groups a common benchmark for fusion-based traversability methods.
  • The targetless calibration procedure means a robot can re-align its thermal and RGB cameras in the field after off-road shocks, instead of returning to a calibration target.
  • Because the pseudo-label is normalized by speed, the resulting costmaps are momentum-aware: obstacles that are low-risk at slow speed are marked as more hazardous at higher speed.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The pseudo-label could be extended with lateral acceleration, wheel slip, or traction estimates, which would capture loss-of-traction events that vertical vibration alone misses; this is an inference, not a claim of the paper.
  • The LiDAR-bridged calibration idea may generalize to other sensor pairs whose appearances differ but which share structural edges, such as radar-to-camera registration.
  • If quantitative field trials confirm the qualitative costmap gains, thermal cameras could become a practical substitute for active illumination in night-time off-road perception.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes IRisPath, a fusion architecture that combines RGB and LWIR imagery with ego-velocity to produce self-supervised traversability costmaps for off-road navigation. It also introduces a day/night off-road dataset collected with the Copernicus vehicle and a targetless extrinsic calibration procedure that uses LiDAR intensity images to align RGB, LWIR, and LiDAR. The main experimental claim is that the fused RGB+IR costmaps are more accurate than single-modality costmaps in day and night conditions, but this claim is supported only by qualitative costmap figures. A secondary claim is that the targetless calibration achieves translation error of 1.7 cm and rotation error of 0.827 degrees, which is not consistently supported by the per-axis table.

Significance. If established, the approach would be practically valuable: the dataset appears to be one of the first off-road collections with co-registered LWIR and RGB day/night data, and the targetless calibration could ease field deployment by avoiding specialized calibration targets. The authors also commit to releasing code and data, which would benefit the community. However, the central traversability claim is currently not supported by any quantitative evaluation, and the self-supervised cost labels are not validated against an independent measure of terrain difficulty. The contribution is therefore best assessed as a promising system description and dataset release, pending rigorous experimental support.

major comments (5)
  1. [Sec. IV.D, Figs. 8 and 9] The central claim that fusion "significantly improves" traversability is supported only by qualitative costmap visualizations. There is no quantitative metric (RMSE, MAE, rank correlation, precision/recall for high-cost regions), no held-out test split, no comparison to TerraPN or other baselines, no error bars, and no statistical test. The text describes differences in color and boundary sharpness, but color scaling and the selection of single frames can make arbitrary differences appear meaningful. Without a quantitative comparison of fused versus single-modality predictions, contribution (4) and the abstract's central claim are unsubstantiated. Please add a quantitative evaluation on a held-out split with standard traversability metrics.
  2. [Sec. III.C.3, Eq. (5)] The cost label is defined as PSD(acc_z) / sqrt(Vx^2 + Vy^2 + 10), but the paper never validates this quantity as a true measure of traversability against independent ground truth. Because both training and the qualitative "improvement" evaluation use this same self-supervised label, even a quantitative comparison on this label would not establish that the model understands terrain difficulty. The paper should at least report correlation with manually annotated terrain difficulty or with a separate vehicle-response measure (e.g., wheel slip or vibration measured at a different location), and it should justify the PSD window length, frequency band, and the constants in Eq. (5).
  3. [Table III and Abstract] The abstract reports rotation accuracy of ±0.827 degrees, but Table III lists per-axis errors of roll 0.2°, pitch 2.11°, and yaw 0.171°; 0.827 is only the mean of these three absolute values, and reporting the mean hides the fact that pitch error is 2.11°, which is likely significant for image projection. The translation error is similarly presented as a single mean without the per-axis distribution. In addition, no repeated trials, standard deviations, or accuracy of the lab measurement reference are reported. The calibration accuracy claim should be restated with the full error distribution, including the worst-axis error and the uncertainty of the reference measurement.
  4. [Sec. III.C and IV] Several quantities needed to reproduce the model are missing: the Fourier-feature standard deviation σ in Eq. (4), patch size i and stride s in Eq. (3), the PSD window length and frequency band used in Eq. (5), and all training hyperparameters (learning rate, epochs, train/validation split, augmentation, optimizer). Since the authors promise to release code and data, these details should be included in the paper or a linked technical appendix to allow independent verification of the central claim.
  5. [Sec. III.B and IV.B] The calibration method is not compared with any existing targetless or target-based calibration baseline, and the choice of feature matcher (SuperGlue versus ORB) is not ablated. Because "novel targetless calibration" is one of the three stated contributions, a quantitative comparison against at least one prior method is needed to support the novelty and accuracy claims; Table III alone, against an unspecified physical measurement tool, is insufficient.
minor comments (6)
  1. [Abstract] The abstract should state that the reported calibration errors are mean absolute errors and should also give the maximum per-axis error, since the current wording implies a guaranteed bound.
  2. [Eq. (3)] The formula for numImagePairs has ambiguous formatting; it should be written as floor((w - i)/s + 1) * floor((h - i)/s + 1).
  3. [Table III] The error sign convention is not defined; specify whether error is measured minus estimate or estimate minus measured.
  4. [Fig. 1 caption] The caption contains a typo: "poincloud" should be "point cloud."
  5. [Sec. III.C.1] The text says that training patches are extracted only beneath the robot, while test-time costmaps use a patch grid over the entire image; please clarify how the model trained on underfoot patches generalizes to the full image and how the front-view patches are labeled.
  6. [Abstract and Sec. IV.A] The abstract calls the dataset labels "pseudo-labels," but Section III.C.3 describes them as self-supervised training labels from Eq. (5); please clarify whether these are the same quantity and how the pseudo-labels are post-processed for the released dataset.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the core pipeline is standard self-supervised regression with an external calibration check, and the main weakness is a lack of quantitative validation rather than a derivation that reduces to its own inputs.

full rationale

The paper does not contain a claimed derivation that reduces to its own inputs by construction. The calibration result, rgbTir = rgbTli (irTli)^-1 (Eq. 2), is a standard composition of independently estimated transforms, and its accuracy is checked against independent lab measurements reported in Table III, with signed errors listed for translation and rotation. The traversability model is trained self-supervised using the IMU power-spectral-density cost of Eq. (5); this defines the training target but does not make the prediction tautological, because at inference the model is applied to sub-images across the full field of view, including regions that were not directly under the robot at training time, and the claimed fusion advantage is presented as a qualitative comparison of predicted costmaps. Whether Eq. (5) is a faithful measure of true terrain difficulty is a legitimate validity concern, and the absence of a quantitative held-out evaluation weakens the evidential support for the central claim, but neither constitutes a circularity pattern of the kind enumerated here: there is no self-definition, no fitted parameter renamed as a prediction, no load-bearing self-citation chain, no uniqueness theorem imported from the authors' prior work, and no ansatz smuggled in via citation. Citations to prior work such as TerraPN and Fourier features are used for architectural inspiration and are not invoked to force the paper's conclusions. The paper's main limitation is evidential, not circular. Therefore the circularity score is 0.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

No new physical entities are introduced. IRisPath is a network architecture and the dataset is a collected artifact, not an invented entity. The free parameters are mostly unreported hyperparameters that the central claim depends on.

free parameters (4)
  • Sigma of Fourier feature B distribution = not reported (tuned)
    Eq. (4), tuned via hyperparameter search; affects velocity encoding and thus the costmap.
  • Constant 10 in cost label denominator = 10
    Eq. (5), chosen to prevent division by near-zero velocities; arbitrary and no sensitivity analysis.
  • Patch size i and stride s = not reported
    Eq. (3), chosen to set output resolution and compute; actual values are not given anywhere in the paper.
  • PSD window length and frequency band = not reported
    The cost label y in Eq. (5) depends on PSD(acc_z), but the window length, overlap, and frequency range are unspecified.
assumptions (4)
  • domain assumption PSD of z-axis acceleration normalized by horizontal velocity is a valid proxy for terrain traversability.
    Section III.C.3: the entire self-supervised label is built on this premise.
  • domain assumption LiDAR intensity image shares matchable features with both RGB and IR images for SuperGlue/ORB.
    Section III.B: the calibration pipeline depends on this feature correspondence.
  • domain assumption The RGB camera, IR camera, and LiDAR have a shared field of view.
    Section III.B: stated as a requirement for the calibration procedure.
  • standard math RANSAC and reprojection error minimization yield the extrinsic transforms.
    Section III.B, Eq. (1): standard estimation procedure, but the equation as written has a frame error.

how reviews work

0 comments
Cite this review

Pith. "Pith review of IRisPath: Enhancing Costmap for Off-Road Navigation with Robust IR-RGB Fusion for Improved Day and Night Traversability." pith.science (2026). https://pith.science/paper/LNAF4SAN

@misc{pith2026241203173,
  author       = {Pith},
  title        = {Pith review of: IRisPath: Enhancing Costmap for Off-Road Navigation with Robust IR-RGB Fusion for Improved Day and Night Traversability},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LNAF4SAN}},
  note         = {Machine review of arXiv:2412.03173}
}
read the original abstract

Autonomous off-road navigation is required for applications in agriculture, construction, search and rescue and defence. Traditional on-road autonomous methods struggle with dynamic terrains, leading to poor vehicle control in off-road conditions. Recent deep-learning models have used perception sensors along with kinesthetic feedback for navigation on such terrains. However, this approach has out-of-domain uncertainty. Factors like change in time of day and weather impacts the performance of the model. We propose a multi modal fusion network "IRisPath" capable of using Thermal and RGB images to provide robustness against dynamic weather and light conditions. To aid further works in this domain, we also open-source a day-night dataset with Thermal and RGB images along with pseudo-labels for traversability. In order to co-register for fusion model we also develop a novel method for targetless extrinsic calibration of Thermal, LiDAR and RGB cameras with translation accuracy of +/-1.7cm and rotation accuracy of +/-0.827degrees.

Figures

Figures reproduced from arXiv: 2412.03173 by the authors.

Figure 1
Figure 1. A rich multi-modal dataset being collected using Copernicus in an off-road terrain during day and night time for learning traversability costmap of the terrain. Dataset contains LWIR and RGB images along with poincloud from LiDARs and stereo camera. Abstract— Autonomous off-road navigation is required for applications in agriculture, construction, search and rescue and defence. Traditional on-road autonomous methods… view at source ↗
Figure 3
Figure 3. IR-RGB Calibration Process to obtain extrinsic transforma￾tion parameters B. Calibration To integrate both IR and RGB modalities in our model, extrinsic calibration between the two cameras is essential. Co-registering IR and RGB images is challenging, as it typically requires specialized checkerboards. In off-road en￾vironments, sensor displacement is common, necessitating frequent recalibration. Target-based method… view at source ↗
Figure 2
Figure 2. Various sensors mounted on Copernicus for the off-road dataset collection. We employ a customized version of the all-terrain ground robot Copernicus, developed by BotSync, to collect the data as seen in [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Overview of the proposed early fusion approach. The inputs to the model are in the purple boxes. To get both the input image in the same frame we apply homographic projection onto the RGB image so that we get matching image pairs of RGB and IR. Then we first split both…
Figure 5
Figure 5. Figure 5: System configuration for dataset collection on the Coperni￾cus vehicle. We initially record all the sensors and actuators as a rosbag and further process them to other formats. B. Extrinsic Calibration To evaluate our calibration technique, we collected data from vario…
Figure 6
Figure 6. Figure 6: The RGB image is being transformed into the IR image frame using the calculated extrinsic parameters [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 8
Figure 8. Figure 8: Costmaps generated by using only single modality in day and night time (a) Day time inference using fusion model (b) Night time inference using fusion mode [PITH_FULL_IMAGE:figures/full_fig_p006_8.png]
Figure 7
Figure 7. Figure 7: Qualitative analysis of the model’s output with respect to the input velocity and fourier features used in the model. study. The costmaps are color-coded, with higher traversabil￾ity costs shown in yellow and lower costs in blue. The results are as follows: 1) Effect o…
Figure 9
Figure 9. Figure 9: Costmaps using fused modality in day and night time. defines boundaries more precisely than IR-only costmaps. At night ( [PITH_FULL_IMAGE:figures/full_fig_p006_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 27 canonical work pages

  1. [1]

    Off-road obstacle avoidance through end-to-end learning

    Yann LeCun, Urs Muller, Jan Ben, Eric Cosatto, and Beat Flepp. Off-road obstacle avoidance through end-to-end learning. In Neural Information Processing Systems , 2005

    work page 2005

  2. [2]

    Rehg, and A.F

    Dongshin Kim, Jie Sun, Sang Min Oh, J.M. Rehg, and A.F. Bobick. Traversability classification using unsupervised on-line visual learning for outdoor robot navigation. In Proceedings 2006 IEEE International Conference on Robotics and Automation, 2006. ICRA 2006. , pages 518–525, 2006

    work page 2006

  3. [3]

    Terrapn: Unstructured terrain navigation using online self-supervised learning

    Adarsh Jagan Sathyamoorthy, Kasun Weerakoon, Tianrui Guan, Jing Liang, and Dinesh Manocha. Terrapn: Unstructured terrain navigation using online self-supervised learning. In 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages 7197– 7204, 2022

    work page 2022

  4. [4]

    Offseg: A semantic segmentation framework for off-road driving

    Kasi Viswanath, Kartikeya Singh, Peng Jiang, Sujit P.B., and Srikanth Saripalli. Offseg: A semantic segmentation framework for off-road driving. 2021 IEEE 17th International Conference on Automation Science and Engineering (CASE) , pages 354–359, 2021

    work page 2021

  5. [5]

    Ttm: Terrain traversability mapping for autonomous excavator navi- gation in unstructured environments

    Tianrui Guan, Zhenpeng He, Dinesh Manocha, and Liangjun Zhang. Ttm: Terrain traversability mapping for autonomous excavator navi- gation in unstructured environments. ArXiv, abs/2109.06250, 2021

    arXiv 2021

  6. [6]

    Ga-nav: Efficient terrain segmentation for robot navigation in unstructured out- door environments

    Tianrui Guan, Divya Kothandaraman, Rohan Chandra, Adarsh Jagan Sathyamoorthy, Kasun Weerakoon, and Dinesh Manocha. Ga-nav: Efficient terrain segmentation for robot navigation in unstructured out- door environments. IEEE Robotics and Automation Letters , 7:8138– 8145, 2021

    work page 2021

  7. [7]

    A Robust Approach to High-Speed Navigation for Unrehearsed Desert Terrain , pages 45–102

    Chris Urmson, Joshua Anhalt, Daniel Bartz, Michael Clark, Tu- grul Galatali, Alexander Gutierrez, Sam Harbaugh, Josh Johnston, Hiroki “Yu” Kato, Phillip Koon, William Messner, Nick Miller, Aaron Mosher, Kevin Peterson, Charlie Ragusa, David Ray, Bryon Smith, Jarrod Snider, Spencer Spiker, Josh Struble, Jason Ziglar, and William “Red” Whittaker. A Robust A...

    work page 2007

  8. [8]

    Semantic terrain classification for off-road autonomous driving

    Amirreza Shaban, Xiangyun Meng, JoonHo Lee, Byron Boots, and Dieter Fox. Semantic terrain classification for off-road autonomous driving. In Aleksandra Faust, David Hsu, and Gerhard Neumann, editors, Proceedings of the 5th Conference on Robot Learning , volume 164 of Proceedings of Machine Learning Research , pages 619–629. PMLR, 08–11 Nov 2022

    work page 2022

Show all 35 references
  1. [9]

    Gregory, Felix Sanchez, John G

    Mateo Guaman Castro, Samuel Triest, Wenshan Wang, Jason M. Gregory, Felix Sanchez, John G. Rogers, and Sebastian Scherer. How does it feel? self-supervised costmap learning for off-road vehicle traversability. In 2023 IEEE International Conference on Robotics and Automation (I...

    2023

  2. [10]

    Wigness, Sungmin Eum, John G

    Maggie B. Wigness, Sungmin Eum, John G. Rogers, David Han, and Heesung Kwon. A rugd dataset for autonomous navigation and visual perception in unstructured outdoor environments. 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages 5000–5007, 2019

    2019

  3. [11]

    Osteen, Maggie B

    Peng Jiang, Philip R. Osteen, Maggie B. Wigness, and Srikanth Saripalli. Rellis-3d dataset: Data, benchmarks and analysis. 2021 IEEE International Conference on Robotics and Automation (ICRA) , pages 1110–1116, 2020

    2021

  4. [12]

    Oliveira, Thomas Brox, and Wolfram Burgard

    Abhinav Valada, Gabriel L. Oliveira, Thomas Brox, and Wolfram Burgard. Deep multispectral semantic scene understanding of forested environments using multimodal fusion. In International Symposium on Experimental Robotics , 2016

    2016

  5. [13]

    The goose dataset for perception in unstructured environments

    Peter Mortimer, Raphael Hagmanns, Miguel Granero, Thorsten Luet- tel, Janko Petereit, and Hans-Joachim Wuensche. The goose dataset for perception in unstructured environments. 2024 IEEE International Conference on Robotics and Automation (ICRA) , pages 14838–14844

    2024

  6. [14]

    Suominen, Jari Yli-Hietanen, Sari Peltonen, Jussi Collin, and Atanas P

    Ihtisham Ali, Ahmed Durmush, Olli J. Suominen, Jari Yli-Hietanen, Sari Peltonen, Jussi Collin, and Atanas P. Gotchev. Finnforest dataset: A forest landscape for visual slam. Robotics Auton. Syst. , 132:103610

  7. [15]

    Barfoot, and Franccois Pomerleau

    Matvej Boxan, Alexander Krawciw, Effie Daum, Xinyuan Qiao, Sven Lilge, Tim D. Barfoot, and Franccois Pomerleau. Fomo: A proposal for a multi-season dataset for robot navigation in for ˆet montmorency. ArXiv, abs/2404.13166, 2024

    2024 arXiv

  8. [16]

    Tartandrive 2.0: More modalities and better infrastructure to further self-supervised learning research in off- road driving tasks

    Matthew Sivaprakasam, Parv Maheshwari, Mateo Guaman Castro, Samuel Triest, Micah Nye, Steven Willits, Andrew Saba, Wenshan Wang, and Sebastian Scherer. Tartandrive 2.0: More modalities and better infrastructure to further self-supervised learning research in off- road driving ...

    2024

  9. [17]

    Jackel, Eric Krotkov, Michael Perschbacher, James Pippine, and Chad Sullivan

    Larry D. Jackel, Eric Krotkov, Michael Perschbacher, James Pippine, and Chad Sullivan. The darpa lagr program: Goals, challenges, methodology, and phase i results. Journal of Field Robotics , 23, 2006

    2006

  10. [18]

    Addison, W

    Reinhold Behringer, Shankar Sundareswaran, Brian Gregory, Richard E Elsley, B. Addison, W. A. Guthmiller, Robert Daily, and David M. Bevly. The darpa grand challenge - development of an autonomous vehicle. IEEE Intelligent V ehicles Symposium, 2004 , pages 226–231, 2004

    2004

  11. [19]

    Where should i walk? predicting terrain properties from images via self-supervised learning

    Lorenz Wellhausen, Alexey Dosovitskiy, Ren ´e Ranftl, Krzysztof Walas, C ´esar Cadena, and Marco Hutter. Where should i walk? predicting terrain properties from images via self-supervised learning. IEEE Robotics and Automation Letters , 4:1509–1516, 2019

    2019

  12. [20]

    Dario Bellicoso, Takahiro Miki, and Marco Hutter

    Peter Fankhauser, Marko Bjelonic, C. Dario Bellicoso, Takahiro Miki, and Marco Hutter. Robust rough-terrain locomotion with a quadrupedal robot. In 2018 IEEE International Conference on Robotics and Automation (ICRA) , pages 5761–5768, 2018

    2018

  13. [21]

    Target-aware dual adversarial learning and a multi-scenario multi-modality benchmark to fuse infrared and visible for object detection

    Jinyuan Liu, Xin Fan, Zhanbo Huang, Guanyao Wu, Risheng Liu, Wei Zhong, and Zhongxuan Luo. Target-aware dual adversarial learning and a multi-scenario multi-modality benchmark to fuse infrared and visible for object detection. In 2022 IEEE/CVF Conference on Computer Vision and...

    2022

  14. [22]

    Madhava Krishna

    Ankit Dhall, Kunal Chelani, Vishnu Radhakrishnan, and K. Madhava Krishna. Lidar-camera calibration using 3d-3d point correspondences. ArXiv, abs/1705.09785, 2017

    2017 arXiv

  15. [23]

    Single-shot is enough: Panoramic infrastructure based calibration of multiple cameras and 3d lidars

    Chuangjie Fang, Shuai Ding, Zilong Dong, Honghua Li, Siyu Zhu, and Ping Tan. Single-shot is enough: Panoramic infrastructure based calibration of multiple cameras and 3d lidars. 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages 8890–8897, 2021

    2021

  16. [24]

    Lidar and camera calibration using motions estimated by sensor fusion odometry

    Ryoichi Ishikawa, Takeshi Oishi, and Katsushi Ikeuchi. Lidar and camera calibration using motions estimated by sensor fusion odometry. 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 7342–7349, 2018

    2018

  17. [25]

    Calib-anything: Zero- training lidar-camera extrinsic calibration method using segment any- thing

    Zhaotong Luo, Guohang Yan, and Yikang Li. Calib-anything: Zero- training lidar-camera extrinsic calibration method using segment any- thing. ArXiv, abs/2306.02656, 2023

    2023 arXiv

  18. [26]

    Mfcalib: Single-shot and automatic extrinsic calibration for lidar and camera in targetless environments based on multi-feature edge

    Tianyong Ye, Wei Xu, Chunran Zheng, and Yukang Cui. Mfcalib: Single-shot and automatic extrinsic calibration for lidar and camera in targetless environments based on multi-feature edge. 2024

    2024

  19. [27]

    Spatial calibration for thermal-rgb cameras and inertial sensor system

    Yan Li, Jindong Tan, Yinlong Zhang, Wei Liang, and Hongsheng He. Spatial calibration for thermal-rgb cameras and inertial sensor system. 2018 24th International Conference on Pattern Recognition (ICPR) , pages 2295–2300, 2018

    2018

  20. [28]

    Abu-Nabah, Mohammad O

    Ahmed ElSheikh, Bassam A. Abu-Nabah, Mohammad O. Hamdan, and Gui-Yun Tian. Infrared camera geometric calibration: A review and a precise thermal radiation checkerboard target. Sensors, 23(7)

  21. [29]

    Tai Fu, Huai Yu, Yaoyu Hu, and Sebastian A. Scherer. Targetless extrinsic calibration of stereo cameras, thermal cameras, and laser sensors in the wild. ArXiv, abs/2109.13414, 2021

    2021 arXiv

  22. [30]

    Srinivasan, Ben Mildenhall, Sara Fridovich-Keil, Nithin Raghavan, Utkarsh Singhal, Ravi Ramamoorthi, Jonathan T

    Matthew Tancik, Pratul P. Srinivasan, Ben Mildenhall, Sara Fridovich-Keil, Nithin Raghavan, Utkarsh Singhal, Ravi Ramamoorthi, Jonathan T. Barron, and Ren Ng. Fourier features let networks learn high frequency functions in low dimensional domains. ArXiv, abs/2006.10739, 2020

    2006 arXiv

  23. [31]

    Fast-lio2: Fast direct lidar-inertial odometry

    Wei Xu, Yixi Cai, Dongjiao He, Jiarong Lin, and Fu Zhang. Fast-lio2: Fast direct lidar-inertial odometry. IEEE Transactions on Robotics , 38:2053–2073, 2021

    2021

  24. [32]

    Lio-sam: Tightly-coupled lidar inertial odometry via smoothing and mapping

    Tixiao Shan, Brendan Englot, Drew Meyers, Wei Wang, Carlo Ratti, and Daniela Rus. Lio-sam: Tightly-coupled lidar inertial odometry via smoothing and mapping. In 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages 5135–5142, 2020

    2020

  25. [33]

    Vision meets robotics: The kitti dataset

    Andreas Geiger, Philip Lenz, Christoph Stiller, and Raquel Urtasun. Vision meets robotics: The kitti dataset. International Journal of Robotics Research (IJRR) , 2013

    2013

  26. [34]

    Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom

    Holger Caesar, Varun Bankiti, Alex H. Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom. nuscenes: A multimodal dataset for autonomous driving. In CVPR, 2020

    2020

  27. [35]

    Superglue: Learning feature matching with graph neural networks

    Paul-Edouard Sarlin, Daniel DeTone, Tomasz Malisiewicz, and An- drew Rabinovich. Superglue: Learning feature matching with graph neural networks. In 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 4937–4946, 2020

    2020

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.