Pith. sign in

REVIEW 4 major objections 4 minor 39 references

RSRNav: Reasoning Spatial Relationship for Image-Goal Navigation

T0 review · 4 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read RSRNav claims that feeding a policy network correlations between goal and observation images, rather than concatenated semantic features, outperforms prior image-goal navigation methods, especially when the goal image comes from a…

desk verdict Solid, well-ablated ImageNav method with a genuinely different perceptual cue; the headline user-matched results are plausible but rest on an untested zero-shot camera transfer, so treat them as provisional. read the letter →

arxiv 2504.17991 v2 pith:UJQXN73I submitted 2025-04-25 cs.CV cs.RO

Zheng Qin , Le Wang , Yabing Wang , Sanping Zhou , Gang Hua , Wei Tang This is my paper · ORCID
classification cs.CVcs.RO
keywords image-goalnavigationspatialrelationshipreasoningvisualcorrelationdirection-awareuser-matchedgoalreinforcementlearningembodiedcross-domaingeneralization
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

This paper aims to establish that an image-goal navigation agent should decide where to go from an explicit spatial-relationship signal between the goal image and the agent's current view, rather than from concatenated semantic feature vectors. The proposed method, RSRNav, computes similarity correlations between goal and observation features and feeds only those correlations to a reinforcement-learned policy; the authors report that even a two-number correlation cue beats the full semantic-feature baseline. Enriching the cue to a dense all-pairs cross-correlation and then to a direction-aware multi-scale correlation pyramid produces large gains in navigation efficiency (Success weighted by Path Length) and in final pose accuracy. The paper reports that RSRNav surpasses previous state-of-the-art methods on all metrics in the user-matched goal setting, where the goal photo is taken from a different camera height, pitch, and field of view than the agent's own camera, which is the setting closest to how a real user would supply a goal image.

What carries the argument

The load-bearing object is the correlation cue $C_{\mathrm{cue}}$, a tensor of feature-similarity scores between the goal image and the current egocentric image. A weight-sharing ResNet-9 encodes both images; the Minimalist version computes two dot-product scores from left and right feature vectors, the Dense version forms the all-pairs matrix $C(i,j)=\langle F_g(i),F_t(j)\rangle$ and flattens it, and the Direction-aware version builds a correlation pyramid $C_s$ by pooling $C$ over the goal-feature dimensions and applies a center-based lookup $P(x)_s = \{x/2^s + r : \|r\|\le 1\}$, yielding a 32-dimensional direction vector per observation point at each scale. These vectors are concatenated across scales, passed through convolutional layers, and flattened to produce the final policy input. The machinery converts the vague notion of 'where is the goal relative to me' into a fixed-size tensor that the policy network can read directly, and the ablations attribute the reported performance gains to this cue.

What would settle it

Render goal images on unseen Gibson scenes with pitch deltas of about $\pm 60^\circ$, outside the paper's reported $\mathrm{U}(-45^\circ,45^\circ)$ extreme range, and run the frozen RSRNav policy; if SPL and success rate fall to the level of the semantic-feature baseline, then the direction-aware correlation's viewpoint robustness is limited to the distribution it was trained and tested on.

Watch

Extended reading notes

Core claim

RSRNav's central claim is that the spatial relationship between the goal and the current observation, expressed as visual correlations, is the information that matters most for the navigation policy. The paper constructs this cue in three escalating forms: two normalized dot-product scores over left and right image halves; a flattened dense cross-correlation matrix over all feature-point pairs; and a direction-aware correlation pyramid with a center-based 3x3 lookup per feature position that produces a multi-scale direction vector. With these cues, a simple untrained ResNet-9 backbone outperforms methods built on large pretrained encoders such as CLIP and ViT. On the Gibson test split the final agent reaches 69.5% SPL and 91.1% success under agent-matched goals, and 56.6% SPL and 83.2% success under user-matched goals without any fine-tuning; on MP3D and HM3D the user-matched gains over prior methods are even larger. For the authors, this shows that correlation carries directional and viewpoint-relative information that semantic features lack, and that a perception-relationship-action policy is the right design for image-goal navigation.

Load-bearing premise

The paper's headline user-matched results rest on a domain-transfer assumption: a policy trained only on goal images taken from the agent's own camera will still navigate well when the goal image comes from a camera with different height, pitch, and field of view, since all compared methods are evaluated on those user-matched images without any fine-tuning.

Editorial extensions

If this is right

  • A two-score correlation cue already beats the full semantic-feature baseline, implying that raw semantic features are not necessary for successful image-goal navigation.
  • As correlation information becomes richer from dense cross-correlation to the direction-aware pyramid, the paper reports roughly fourfold success-rate gains over the minimalist version, suggesting that the amount of relational information, not encoder capacity, drives performance.
  • Under user-matched goal images, RSRNav reports 56.6% SPL and 83.2% success on Gibson, exceeding the strongest prior method by about 10 SPL points, and retains the lead on MP3D and HM3D without fine-tuning.
  • The direction-aware component improves the angle-to-goal at stop, meaning the agent is not just near the goal but is facing it, which matters for tasks where the agent must interact after arriving.
  • Because the method uses a small untrained ResNet-9, the reported gains come with the smallest parameter count among compared methods, which favors on-device deployment.

Reading between the lines

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

  • If the correlation cue transfers across camera intrinsics and extrinsics as the user-matched results suggest, the same relational policy could be applied to other vision-goal tasks, such as navigating to an object photographed by a user's phone, because the cue encodes how two views relate rather than absolute appearance.
  • The paper's stated limitation is that all training data is simulated; a natural extension is to pre-train or fine-tune the correlation cue on real-world photo pairs to test whether the viewpoint robustness survives the sim-to-real gap.
  • The Minimalist version's success suggests a cheap diagnostic: navigation performance as a function of correlation granularity could be measured, letting system designers pick the smallest correlation that meets a target SPL instead of always using the full pyramid.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes RSRNav, an image-goal navigation method that replaces semantic feature concatenation with an explicit correlation cue between the goal image and the current egocentric observation. The method is developed in three progressive versions: a two-score "Minimalist Relationship" cue, a dense cross-correlation matrix (Eq. 1), and a direction-aware multi-scale correlation pyramid with a center-based lookup (Eqs. 3-4). The policy is trained end-to-end with PPO using a reward adapted from ZER (Eqs. 6-7). Experiments on Gibson, MP3D, and HM3D compare against prior methods under both "agent-matched goal" and "user-matched goal" camera settings, with ablations over the three correlation variants, feature-map resolution, and pyramid depth. The central claim is that explicit spatial-relationship reasoning, rather than richer semantic features, yields more efficient navigation and, in particular, more robust performance under user-matched goal images with varying height, pitch, and horizontal field of view.

Significance. If the results hold, this is a useful and conceptually clean contribution: a lightweight ResNet-9 without pretraining, with fewer parameters than prior end-to-end methods, matches or exceeds stronger backbones on Gibson and improves user-matched SPL substantially over FGPrompt. The three-stage development from minimalist to dense to direction-aware correlation is pedagogically effective, and Tables IV-VI provide evidence that correlation cues, rather than raw semantic features, drive the policy's behavior. The cross-domain evaluations on MP3D and HM3D and the user-matched camera setting address a practically important generalization problem. However, the headline zero-shot user-matched claim rests on an untested transfer assumption, the paper lacks statistical support for differences that are often small, and the direction-aware component is not cleanly isolated from an increase in feature-map resolution. These issues need to be addressed before the main claims are fully supported.

major comments (4)
  1. [§I, §III-C3, §IV-A, Tables I-III] The claim that spatial relationships are "less affected by inconsistent viewpoints" (Section I) is load-bearing but is asserted rather than derived or isolated experimentally. The direction-aware lookup in Eq. (4) uses fixed feature-map coordinates x/2^s and is not normalized or conditioned on goal camera height, pitch, or HFOV, which in the user-matched setting are sampled from U(0.8,1.5)m, U(-5,5) degrees, and U(60,120) degrees. A goal at a fixed egocentric bearing projects to different pixel locations under different HFOV and pitch, so it is not evident that correlations learned from agent-matched goals transfer to user-matched goals. The paper's own numbers show a drop from 69.5 to 56.6 SPL on Gibson, confirming that the model is not viewpoint-invariant. Please provide per-camera-parameter breakdowns (for example, HFOV bins and pitch bins), an ablation with normalized or camera-conditioned lookup coordinates, or a fine-tuning control to support the zero-shot transfer claim.
  2. [§IV-A, Tables I-VII] All reported results are averages over three random seeds, but no standard deviations, confidence intervals, or significance tests are provided. This matters because several central claims rest on small gaps: the agent-matched Gibson SPL difference of 69.5 vs 66.5 for FGPrompt-EF, the pyramid layer 2 vs layer 3 difference in Table VI (59.4 vs 61.2), and the claimed +8.0 SPL from dense to direction-aware correlation. Please report per-seed results or error bars and, where the evaluation episodes are shared across methods, paired tests such as bootstrap or Wilcoxon signed-rank tests so the reader can distinguish signal from reinforcement-learning noise.
  3. [§IV-C, Table VI] The ablation labeled "Direction-aware Correlation" changes two factors at once relative to "Dense Correlation": it adds the LookUp operation and the correlation pyramid, and it increases the feature-map size from 4x4 to 16x16 (Section IV-A). Table VI shows that LookUp at 4x4 actually degrades performance (50.4 vs 53.2 SPL), so the improvement from 53.2 to 58.0/61.2 SPL could be due largely to the higher-resolution feature maps rather than to the directional lookup. Please add a 16x16 dense-correlation baseline without LookUp or pyramid, and report the pyramid contribution at fixed 16x16 resolution, before attributing the +8.0 SPL improvement to directional information.
  4. [§IV-C, Table IV and Table III captions] The caption of Table IV states "ALL METHODS ARE TRAINED ON GIBSON AND EVALUATED ON MP3D," while the text of Section IV-C says that all ablation studies train on Gibson and reports "on Gibson" results; Table III's caption similarly contains a repeated MP3D sentence for HM3D. Please clarify the evaluation split used for the ablations. If the design choices in Tables V and VI (pyramid depth, feature-map size, lookup size) were selected on the same Gibson test episodes used for the headline comparisons in Table I, this is selection-on-test and should be stated explicitly and corrected with a held-out validation set.
minor comments (4)
  1. [Throughout] There are several typographical errors that should be corrected: "Realtionship" in the Section III-C headings, "FGPropmt" in Section IV-E, "evalution" in Section IV-B, "Represents" in Table IV, and "Succees !" in the Figure 7 caption.
  2. [§III-C3, Eq. (4)] The index order of the 4D correlation tensor is ambiguous. Equation (1) defines C(i,j) = <Fg(i), Ft(j)>, but the text describes P(x)_s as a region for "each position x in Ft." Please define explicitly which tensor axes correspond to Fg and Ft in the LookUp operation and how the cropped 3x3 region is reshaped to the stated 32-dimensional vector.
  3. [§IV-A, Figure 5] Figure 5 is referenced as comparing the number of parameters with SPL, but the printed text does not include axis labels or numeric values for the parameter counts. Please add these to the figure or provide the values in a table.
  4. [§IV-B, Tables II-III] Table II and Table III report cross-domain results on MP3D and HM3D, but the comparison set differs across tables: FGPrompt is listed without its EF/MF suffixes in Tables II-III. Please state explicitly which FGPrompt variant is used in each comparison so the reader can verify that the same baseline configuration is used.

Circularity Check

1 steps flagged · score 2.0 of 10

Mild circularity: angle-to-goal evidence is partly trained by the reward that explicitly rewards angle reduction; the core SPL/SR comparison is independent.

  1. fitted input called prediction [Section III-D (Eqs. 6-7); Section IV-D (Fig. 4 analysis)]
    "Following [7], the reward consists of two sub-reward functions. The first is calculated at each time step in the episode as follows: r_t = r_d(d_t,d_{t-1}) + [d_t≤d_s] r_α(α_t,α_{t-1}) − γ ... The second reward only exists at the end of an episode, as follows: r_final = 5 × ([d_t≤d_s] + [d_t≤d_s and α_t≤α_s]) ... For the second challenge, whether the angle can be adjusted accurately according to the goal can be reflected most intuitively in the 'Angle to Goal' metric ... demonstrates that correlation can help the agent adjust the pose accurately."

    The policy is trained with a reward that explicitly rewards reducing the angle to the goal view (r_α at each step and an additional +5 at stop when α_t≤α_s). The paper then uses the 'Angle to Goal' metric as evidence that the correlation mechanism enables precise angle adjustment. That metric is directly optimized by the training reward, so the behavior is partly trained-in rather than an independent prediction of the correlation design. Because all compared methods share the same reward, the relative ablation comparison remains meaningful; this is a mild circularity in the narrative, not a collapse of the central SPL/SR comparison.

full rationale

The central claim of RSRNav is empirical: a correlation cue computed by Eqs. (1)-(5) from goal and observation features is fed to a PPO-trained policy and evaluated against external baselines (ZER, FGPrompt, OVRL-V2, etc.) on Gibson, MP3D, and HM3D. The correlation construction is fully specified by the paper's equations, and the SPL/SR improvements are not derived from the definition of the correlation cue; they are measured. There is no load-bearing self-citation chain, no uniqueness theorem imported from the authors' prior work, and no fitted parameter being renamed as a prediction. The user-matched zero-shot transfer is a domain-transfer assumption, which is a correctness risk rather than a circularity. The one mild circular step is the angle-to-goal narrative: the reward function explicitly trains angle reduction, and the paper later cites the 'Angle to Goal' metric as evidence for the directional mechanism. This does not undermine the main benchmarking results, so the overall circularity score is low.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central claim rests on three domain assumptions: viewpoint-transfer from agent-matched to user-matched cameras, the validity of the ZER reward, and sim-to-real representativeness. The method adds no invented physical entities. Several hyperparameters (pyramid depth, feature-map resolution, lookup radius) are hand-chosen and tuned on the test episodes, which is the main parametric burden.

free parameters (3)
  • Correlation pyramid depth S = 3
    Table VI ablates 1, 2, and 3 layers and selects 3; the number of layers is a design choice tuned on the Gibson test episodes.
  • Direction-aware feature map size = 16x16
    Table VI shows 16x16 beats 4x4; input feature map resolution for the lookup operator is a hand-chosen design parameter.
  • Lookup search radius (3x3 region) = 3x3 (||r|| <= 1)
    Equation (4) defines the direction-aware searching region; the radius is fixed by hand, not derived.
assumptions (3)
  • domain assumption Training on Gibson with the agent-matched goal camera transfers to user-matched goal cameras at test time without fine-tuning.
    Section IV-A describes the goal camera settings; Section IV-B states all methods are trained under agent-matched and evaluated under user-matched. The transfer is assumed and is the load-bearing premise for the user-matched results.
  • domain assumption The reward from ZER (Equations 6-7), based on distance and angle to the goal view, is a valid training signal for the navigation task.
    The policy is trained with this reward; if the reward does not align with the success metric (1m distance), the learned policy may not optimize the evaluated objective. This follows [7].
  • domain assumption Simulator scenes (Gibson, MP3D, HM3D) are representative enough of real indoor environments for the method to transfer to the real world.
    The paper's own Limitation section states all training data comes from the simulator and there are gaps from the real world, confirming this is an assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RSRNav: Reasoning Spatial Relationship for Image-Goal Navigation." pith.science (2026). https://pith.science/paper/UJQXN73I

@misc{pith2026250417991,
  author       = {Pith},
  title        = {Pith review of: RSRNav: Reasoning Spatial Relationship for Image-Goal Navigation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UJQXN73I}},
  note         = {Machine review of arXiv:2504.17991}
}
read the original abstract

Recent image-goal navigation (ImageNav) methods learn a perception-action policy by separately capturing semantic features of the goal and egocentric images, then passing them to a policy network. However, challenges remain: (1) Semantic features often fail to provide accurate directional information, leading to superfluous actions, and (2) performance drops significantly when viewpoint inconsistencies arise between training and application. To address these challenges, we propose RSRNav, a simple yet effective method that reasons spatial relationships between the goal and current observations as navigation guidance. Specifically, we model the spatial relationship by constructing correlations between the goal and current observations, which are then passed to the policy network for action prediction. These correlations are progressively refined using fine-grained cross-correlation and direction-aware correlation for more precise navigation. Extensive evaluation of RSRNav on three benchmark datasets demonstrates superior navigation performance, particularly in the "user-matched goal" setting, highlighting its potential for real-world applications.

Figures

Figures reproduced from arXiv: 2504.17991 by the authors.

Figure 1
Figure 1. Illustration of human-intuitive navigation, with [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of our RSRNav. “Minimalist Relationship” constructs sparse correlations to initially explore the importance of spatial relationship in ImageNav. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Graphical representation of the correlation pyramid and center-based [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Comparison of convergence rates and tests results on metrics “Distance to Goal” and “Angle to Goal”. The former metric measures the distance [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Compare the number of parameters, as well as the SPL on Gibson. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Comparison of Visualization Results. The gray area in the scene represents the area scanned by the egocentric camera while the agent is traveling. [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Comparison of three methods. The left side shows the target viewpoint image, the middle shows the image captured by the agent’s own camera after [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 28 canonical work pages

  1. [1]

    Memory- adaptive vision-and-language navigation,

    K. He, Y . Jing, Y . Huang, Z. Lu, D. An, and L. Wang, “Memory- adaptive vision-and-language navigation,” Pattern Recognition, vol. 153, p. 110511, 2024

    work page 2024

  2. [2]

    Visual slam for robot navigation in healthcare facility,

    B. Fang, G. Mei, X. Yuan, L. Wang, Z. Wang, and J. Wang, “Visual slam for robot navigation in healthcare facility,” Pattern recognition, vol. 113, p. 107822, 2021

    work page 2021

  3. [3]

    Navcomposer: Composing language instructions for navigation trajectories through action-scene-object modularization,

    Z. He, L. Wang, L. Chen, C. Liu, and Q. Chen, “Navcomposer: Composing language instructions for navigation trajectories through action-scene-object modularization,” IEEE Transactions on Circuits and Systems for Video Technology, 2025

    work page 2025

  4. [4]

    Erd: Encoder-residual-decoder neural network for underwater image enhancement,

    J. Cao, W. Peng, Y . Liu, J. Dong, P. Le Callet, and S. Kwong, “Erd: Encoder-residual-decoder neural network for underwater image enhancement,” IEEE Transactions on Circuits and Systems for Video Technology, 2025

    work page 2025

  5. [5]

    Learning to explore using active neural slam,

    D. S. Chaplot, D. Gandhi, S. Gupta, A. Gupta, and R. Salakhutdinov, “Learning to explore using active neural slam,” in Proc. Int. Conf. Learn. Represent., 2020

    work page 2020

  6. [6]

    Offline visual representation learning for embodied navigation,

    K. Yadav, R. Ramrakhya, A. Majumdar, V .-P. Berges, S. Kuhar, D. Batra, A. Baevski, and O. Maksymets, “Offline visual representation learning for embodied navigation,” in Proc. Int. Conf. Learn. Represent. , 2023

    work page 2023

  7. [7]

    Zero experience required: Plug & play modular transfer learning for semantic visual navigation,

    Z. Al-Halah, S. K. Ramakrishnan, K. Grauman, and K. Grauman, “Zero experience required: Plug & play modular transfer learning for semantic visual navigation,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , 2022, pp. 17 031–17 041

    work page 2022

  8. [8]

    Ovrl-v2: A simple state-of-art baseline for imagenav and objectnav,

    K. Yadav, A. Majumdar, R. Ramrakhya, N. Yokoyama, A. Baevski, Z. Kira, O. Maksymets, and D. Batra, “Ovrl-v2: A simple state-of-art baseline for imagenav and objectnav,” arXiv preprint arXiv:2303.07798, 2023

    arXiv 2023

Show all 39 references
  1. [9]

    Zson: Zero-shot object-goal navigation using multimodal goal embed- dings,

    A. Majumdar, G. Aggarwal, B. Devnani, J. Hoffman, and D. Batra, “Zson: Zero-shot object-goal navigation using multimodal goal embed- dings,” in Proc. Adv. Neural Inf. Process. Syst., 2022, pp. 32 340–32 352

    2022

  2. [10]

    Visual graph memory with unsupervised representation for visual navigation,

    O. Kwon, N. Kim, Y . Choi, H. Yoo, J. Park, and S. Oh, “Visual graph memory with unsupervised representation for visual navigation,” inProc. IEEE Int. Conf. Comput. Vis. , 2021, pp. 15 890–15 899

    2021

  3. [11]

    Memonav: Working memory model for visual navigation,

    H. Li, Z. Wang, X. Yang, Y . Yang, S. Mei, and Z. Zhang, “Memonav: Working memory model for visual navigation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 17 913–17 922

    2024

  4. [12]

    A frontier-based approach for autonomous exploration,

    B. Yamauchi, “A frontier-based approach for autonomous exploration,” in CIRA, 1997, pp. 146–151

    1997

  5. [13]

    Slam2: Simultaneous localization and multimode mapping for indoor dynamic environments,

    Z. Lin, Q. Zhang, Z. Tian, P. Yu, Z. Ye, H. Zhuang, and J. Lan, “Slam2: Simultaneous localization and multimode mapping for indoor dynamic environments,” Pattern Recognition, vol. 158, p. 111054, 2025

    2025

  6. [14]

    Navigating to objects specified by images,

    J. Krantz, T. Gervet, K. Yadav, A. Wang, C. Paxton, R. Mottaghi, D. Batra, J. Malik, S. Lee, and D. S. Chaplot, “Navigating to objects specified by images,” arXiv preprint arXiv:2304.01192 , 2023

    2023 arXiv

  7. [15]

    Habitat-web: Learning embodied object-search strategies from human demonstrations at scale,

    R. Ramrakhya, E. Undersander, D. Batra, and A. Das, “Habitat-web: Learning embodied object-search strategies from human demonstrations at scale,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., 2022, pp. 5173–5183

    2022

  8. [16]

    Last- mile embodied visual navigation,

    J. Wasserman, K. Yadav, G. Chowdhary, A. Gupta, and U. Jain, “Last- mile embodied visual navigation,” in Conference on Robot Learning . PMLR, 2023, pp. 666–678

    2023

  9. [17]

    Neural topological slam for visual navigation,

    D. S. Chaplot, R. Salakhutdinov, A. Gupta, and S. Gupta, “Neural topological slam for visual navigation,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., 2020, pp. 12 875–12 884

    2020

  10. [18]

    No rl, no simulation: Learning to navigate without navigating,

    M. Hahn, D. S. Chaplot, S. Tulsiani, M. Mukadam, J. M. Rehg, and A. Gupta, “No rl, no simulation: Learning to navigate without navigating,” in Proc. Adv. Neural Inf. Process. Syst. , 2021, pp. 26 661– 26 673

    2021

  11. [19]

    Feudal networks for visual navigation,

    F. Johnson, B. B. Cao, A. Ashok, S. Jain, and K. Dana, “Feudal networks for visual navigation,” arXiv preprint arXiv:2402.12498 , 2024. IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY , VOL. X, NO. X 11

    2024 arXiv

  12. [20]

    Superglue: Learning feature matching with graph neural networks,

    P.-E. Sarlin, D. DeTone, T. Malisiewicz, and A. Rabinovich, “Superglue: Learning feature matching with graph neural networks,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , 2020, pp. 4938–4947

    2020

  13. [21]

    Thda: Treasure hunt data augmentation for semantic navigation,

    O. Maksymets, V . Cartillier, A. Gokaslan, E. Wijmans, W. Galuba, S. Lee, and D. Batra, “Thda: Treasure hunt data augmentation for semantic navigation,” in Proc. IEEE Int. Conf. Comput. Vis. , 2021, pp. 15 374–15 383

    2021

  14. [22]

    Target-driven visual navigation in indoor scenes using deep reinforcement learning,

    Y . Zhu, R. Mottaghi, E. Kolve, J. J. Lim, A. Gupta, L. Fei-Fei, and A. Farhadi, “Target-driven visual navigation in indoor scenes using deep reinforcement learning,” in 2017 IEEE international conference on robotics and automation (ICRA) , 2017, pp. 3357–3364

    2017

  15. [23]

    Goal-conditioned imitation learning,

    Y . Ding, C. Florensa, P. Abbeel, and M. Phielipp, “Goal-conditioned imitation learning,” in Proc. Adv. Neural Inf. Process. Syst. , vol. 32, 2019

    2019

  16. [24]

    Curious representation learning for embodied intelligence,

    Y . Du, C. Gan, and P. Isola, “Curious representation learning for embodied intelligence,” in Proc. IEEE Int. Conf. Comput. Vis. , 2021, pp. 10 408–10 417

    2021

  17. [25]

    Dual-view curricular optimal transport for cross-lingual cross-modal retrieval,

    Y . Wang, S. Wang, H. Luo, J. Dong, F. Wang, M. Han, X. Wang, and M. Wang, “Dual-view curricular optimal transport for cross-lingual cross-modal retrieval,” IEEE Transactions on Image Processing, vol. 33, pp. 1522–1533, 2024

    2024

  18. [26]

    Reading-strategy inspired visual representation learning for text-to- video retrieval,

    J. Dong, Y . Wang, X. Chen, X. Qu, X. Li, Y . He, and X. Wang, “Reading-strategy inspired visual representation learning for text-to- video retrieval,” IEEE Transactions on Circuits and Systems for Video Technology, 2022

    2022

  19. [27]

    Denoising reuse: Exploiting inter-frame motion consistency for efficient video generation,

    C. Wang, S. Yan, Y . Chen, X. Wang, Y . Wang, M. Dong, X. Yang, D. Li, R. Zhu, D. A. Clifton et al. , “Denoising reuse: Exploiting inter-frame motion consistency for efficient video generation,” IEEE Transactions on Circuits and Systems for Video Technology , 2025

    2025

  20. [28]

    Up-person: Unified parameter-efficient transfer learning for text-based person retrieval,

    Y . Liu, Y . Li, X. Lan, W. Yang, Z. Liu, and Q. Liao, “Up-person: Unified parameter-efficient transfer learning for text-based person retrieval,” IEEE Transactions on Circuits and Systems for Video Technology, 2025

    2025

  21. [29]

    FGPrompt: Fine- grained goal prompting for image-goal navigation,

    X. Sun, P. Chen, J. Fan, J. Chen, T. Li, and M. Tan, “FGPrompt: Fine- grained goal prompting for image-goal navigation,” in Proc. Adv. Neural Inf. Process. Syst. , 2024

    2024

  22. [30]

    Prox- imal policy optimization algorithms,

    J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, “Prox- imal policy optimization algorithms,” arXiv preprint arXiv:1707.06347, 2017

    2017 arXiv

  23. [31]

    Memory-augmented reinforcement learning for image-goal navigation,

    L. Mezghan, S. Sukhbaatar, T. Lavril, O. Maksymets, D. Batra, P. Bo- janowski, and K. Alahari, “Memory-augmented reinforcement learning for image-goal navigation,” in IROS, 2022, pp. 3316–3323

    2022

  24. [32]

    Topological semantic graph memory for image-goal navigation,

    N. Kim, O. Kwon, H. Yoo, Y . Choi, J. Park, and S. Oh, “Topological semantic graph memory for image-goal navigation,” in Conference on Robot Learning, 2023, pp. 393–402

    2023

  25. [33]

    On evalu- ation of embodied navigation agents,

    P. Anderson, A. Chang, D. S. Chaplot, A. Dosovitskiy, S. Gupta, V . Koltun, J. Kosecka, J. Malik, R. Mottaghi, M. Savvaet al., “On evalu- ation of embodied navigation agents,” arXiv preprint arXiv:1807.06757, 2018

    2018 arXiv

  26. [34]

    Instance- specific image goal navigation: Training embodied agents to find object instances,

    J. Krantz, S. Lee, J. Malik, D. Batra, and D. S. Chaplot, “Instance- specific image goal navigation: Training embodied agents to find object instances,” arXiv preprint arXiv:2211.15876 , 2022

    2022 arXiv

  27. [35]

    Habitat: A platform for embodied ai research,

    M. Savva, A. Kadian, O. Maksymets, Y . Zhao, E. Wijmans, B. Jain, J. Straub, J. Liu, V . Koltun, J. Malik et al. , “Habitat: A platform for embodied ai research,” in Proc. IEEE Int. Conf. Comput. Vis. , 2019, pp. 9339–9347

    2019

  28. [36]

    Gibson env: Real-world perception for embodied agents,

    F. Xia, A. R. Zamir, Z. He, A. Sax, J. Malik, and S. Savarese, “Gibson env: Real-world perception for embodied agents,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , 2018, pp. 9068–9079

    2018

  29. [37]

    Matterport3d: Learning from rgb-d data in indoor environments,

    A. Chang, A. Dai, T. Funkhouser, M. Halber, M. Niessner, M. Savva, S. Song, A. Zeng, and Y . Zhang, “Matterport3d: Learning from rgb-d data in indoor environments,” International Conference on 3D Vision , 2017

    2017

  30. [38]

    Habitat-matterport 3d dataset (hm3d): 1000 large-scale 3d environments for embodied ai,

    S. K. Ramakrishnan, A. Gokaslan, E. Wijmans, O. Maksymets, A. Clegg, J. Turner, E. Undersander, W. Galuba, A. Westbury, A. X. Chang et al. , “Habitat-matterport 3d dataset (hm3d): 1000 large-scale 3d environments for embodied ai,” in Proc. Adv. Neural Inf. Process. Syst., 2021

    2021

  31. [39]

    Pytorch: An imperative style, high-performance deep learning library,

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga et al. , “Pytorch: An imperative style, high-performance deep learning library,” in Proc. Adv. Neural Inf. Process. Syst. , 2019. Zheng Qin received the B.S. degree in...

    2019

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.