{"id":"30dbe888-21fb-4545-95e3-bc720d6793a1","arxiv_id":"2605.31576","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Joint multi-camera LiDAR extrinsic calibration using learned pairwise initialization followed by geometric bundle adjustment refinement.","lead":"The paper describes a two-stage framework for calibrating multiple cameras with one LiDAR by first getting initial guesses from a learned model for each pair and then refining them all together using geometric optimization. A smart generalist might read it to see how to make sensor setups more consistent for better 3D vision in machines.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Bundle adjustment's ability to enforce global consistency from inconsistent pairwise inits (esp. out-of-domain) depends on untested strength of relative-pose priors vs. data terms.","rationale":"The reader's weakest assumption directly identifies the same load-bearing step. With the full manuscript now available the concern remains unchanged because the abstract-level description already isolates the unverified transfer from pairwise to joint; any additional implementation details would need to be checked against the proposed ablation to confirm or refute the assumption.","tokens_in":1708,"tokens_out":358,"duration_ms":16443,"concrete_test":"On the Walkley sequences, rerun the refinement stage with the relative-pose prior weight set to zero (or reduced by 100x) while keeping reprojection and per-camera terms; measure both per-camera translation error and the variance of inter-camera relative poses before/after. If consistency metrics collapse while per-camera errors stay comparable, the global-consistency claim rests primarily on the priors.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that the multi-frame BA (reprojection + per-camera prior + relative-pose prior) converts independent CMRNext outputs into a single consistent set of camera-LiDAR extrinsics. This is least secure on Walkley, where single-camera errors start at 108.6 cm; if the relative-pose priors are the dominant term or if the optimization landscape contains multiple consistent-but-wrong minima, the reported drop to 3.1 cm could reflect prior regularization rather than geometric coupling. The abstract gives no equation for the relative-pose term, no weight schedule, and no ablation removing it, so it is unclear whether the joint refinement actually resolves inconsistencies or merely averages them under the priors.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript presents a two-stage framework for joint multi-camera LiDAR extrinsic calibration. The first stage applies the CMRNext model independently to each camera-LiDAR pair to obtain initial extrinsics and dense 2D-3D correspondences. The second stage performs a multi-frame bundle adjustment incorporating reprojection error, per-camera prior, and relative-pose prior terms to refine the estimates into a globally consistent calibration. Experiments on the KITTI (in-domain) and Walkley (out-of-domain) datasets report improved per-camera accuracy and inter-camera consistency, with translation errors of 0.89 cm on KITTI and a reduction from 108.6 cm to 3.1 cm on Walkley.","tokens_in":1848,"tokens_out":597,"duration_ms":18340,"significance":"If the central claim holds, the work is significant for addressing inconsistency across independent pairwise camera-LiDAR calibrations in multi-camera rigs, a practical issue in robotics and autonomous driving. The combination of learned pairwise initialization with geometric refinement, plus explicit evaluation on out-of-domain data where single-camera predictions degrade, is a strength. The approach demonstrates that explicit multi-camera coupling can yield large error reductions when independent estimates are unreliable.","major_comments":[{"comment":"Abstract: The central claim that the multi-frame bundle adjustment converts independent CMRNext pairwise predictions into globally consistent extrinsics rests on the relative-pose prior term, yet the abstract provides neither its equation, weighting schedule, nor an ablation removing this term. Without these, it is unclear whether the reported drop from 108.6 cm to 3.1 cm on Walkley reflects geometric coupling or prior regularization, which is load-bearing for the consistency claim.","section":"Abstract"},{"comment":"Method section (bundle adjustment formulation): The assumption that reprojection + per-camera prior + relative-pose prior terms can reliably resolve inconsistencies from independent initializations is untested in the reported experiments; an ablation on the relative-pose prior weight (or its removal) is needed to confirm that the joint optimization enforces consistency rather than averaging under strong priors, especially given the large initial errors on out-of-domain data.","section":"Method"}],"minor_comments":[{"comment":"Abstract: The rotation error of 0.038 is reported without units (radians or degrees), which should be clarified for reproducibility.","section":"Abstract"},{"comment":"Experiments: Baseline comparisons and implementation details (e.g., exact CMRNext usage, optimization hyperparameters) are referenced but not fully detailed in the provided abstract; ensure these are expanded in the full manuscript for clarity.","section":"Experiments"}],"recommendation":"major_revision","confidential_remarks":"The manuscript builds directly on CMRNext; confirm whether this is prior work by the same authors and that self-citation practices are appropriate."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive feedback on our manuscript. We address each major comment below and will incorporate revisions to strengthen the presentation of the relative-pose prior and its empirical validation.","responses":[{"response":"We agree that the abstract should more clearly signal the role of the relative-pose prior. In the revised manuscript we will expand the abstract sentence describing the bundle adjustment to explicitly name the relative-pose prior term and its purpose. Because abstracts have strict length limits, the full equation and weighting schedule will be added to Section 3.2 (Bundle Adjustment Formulation) with a forward reference from the abstract. We will also add the requested ablation (see response to the second comment).","revision_made":"yes","referee_comment":"[Abstract] Abstract: The central claim that the multi-frame bundle adjustment converts independent CMRNext pairwise predictions into globally consistent extrinsics rests on the relative-pose prior term, yet the abstract provides neither its equation, weighting schedule, nor an ablation removing this term. Without these, it is unclear whether the reported drop from 108.6 cm to 3.1 cm on Walkley reflects geometric coupling or prior regularization, which is load-bearing for the consistency claim."},{"response":"We acknowledge that the current experiments do not isolate the contribution of the relative-pose prior. In the revised manuscript we will add a new ablation table (Table X) that reports per-camera and inter-camera translation/rotation errors on both KITTI and Walkley when (i) the relative-pose prior weight is set to zero and (ii) the weight is varied over a range of values. This will directly test whether the observed consistency gains arise from geometric coupling rather than prior regularization alone.","revision_made":"yes","referee_comment":"[Method] Method section (bundle adjustment formulation): The assumption that reprojection + per-camera prior + relative-pose prior terms can reliably resolve inconsistencies from independent initializations is untested in the reported experiments; an ablation on the relative-pose prior weight (or its removal) is needed to confirm that the joint optimization enforces consistency rather than averaging under strong priors, especially given the large initial errors on out-of-domain data."}],"tokens_in":1426,"tokens_out":477,"duration_ms":14434,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The paper takes learned pairwise camera-LiDAR calibration and adds an explicit joint refinement stage to handle the rigid coupling across multiple cameras on one platform. They apply CMRNext independently to each camera for initial extrinsics and dense correspondences, then run a multi-frame bundle adjustment that combines reprojection error with per-camera priors and relative-pose priors between cameras.\n\nThis produces better per-camera accuracy and inter-camera consistency than the independent baselines. On KITTI the numbers are already strong and edge up to 0.89 cm translation and 0.038 rotation error. The clearer win is on the out-of-domain Walkley set, where translation error falls from 108.6 cm to 3.1 cm. That drop indicates the joint stage can rescue results when single-camera predictions are weak.\n\nThe soft spot is the one flagged in the stress-test. The abstract does not give the relative-pose term equation, the weight schedule, or an ablation that removes it. When the starting error is over a meter, it is possible the priors are carrying most of the load and the optimization is simply averaging under regularization rather than enforcing consistency through the data. Without those specifics the central claim that the geometric coupling resolves inconsistencies stays partly untested.\n\nThe rest of the work is standard bundle adjustment on top of an existing learned initializer, so the math is reproducible but not surprising. Citations cover the relevant prior calibration literature.\n\nThis is aimed at researchers who calibrate multi-camera LiDAR rigs for robotics or autonomous driving. A reader working on system-level consistency would get practical value from the reported gains.\n\nIt deserves peer review because the two-stage idea is straightforward, the out-of-domain improvement is sizable, and the remaining questions about the priors are answerable with the full manuscript.","headline":"The joint bundle adjustment after independent learned pairwise inits is the actual addition here, and it shows large gains on out-of-domain data, but whether the relative-pose priors or the geometric coupling drive the consistency is not clear from the given details.","tokens_in":2292,"tokens_out":451,"would_cite":false,"duration_ms":18144,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Pairwise camera-LiDAR predictions refine into a globally consistent multi-camera calibration via joint bundle adjustment.","keywords":["camera-LiDAR calibration","multi-camera systems","extrinsic calibration","bundle adjustment","pairwise initialization","geometric refinement","sensor consistency"],"falsifier":"A multi-camera dataset in which the final inter-camera relative poses after joint refinement deviate farther from ground truth than the original pairwise estimates.","tokens_in":2615,"feed_emoji":"📐","tokens_out":636,"duration_ms":26131,"temperature":0.7,"pith_summary":"Learning-based methods typically calibrate each camera to the LiDAR independently, which can yield accurate per-pair results that still disagree with one another across the full camera set. This paper establishes that feeding those independent predictions into a single geometric refinement step produces extrinsics that remain accurate for every camera while also aligning the cameras with each other. The refinement step runs a multi-frame bundle adjustment that penalizes reprojection error and adds priors on both individual camera poses and their relative positions. The approach matters for any multi-camera platform because downstream perception and mapping tasks require all sensors to share one coherent coordinate frame, and the joint step proves especially helpful when the initial per-camera predictions are unreliable.","feed_headline":"Joint refinement turns pairwise calibrations into consistent multi-camera LiDAR setup","feed_subtitle":"Independent per-camera estimates often disagree; bundle adjustment enforces alignment and cuts errors on out-of-domain data.","key_machinery":"The multi-frame bundle adjustment that jointly optimizes all camera-LiDAR extrinsics from independent pairwise initializations by combining reprojection, per-camera prior, and relative-pose prior terms.","core_discovery":"The central claim is that a two-stage process—independent learned pairwise initialization for each camera-LiDAR pair followed by joint multi-frame bundle adjustment using reprojection error, per-camera prior, and relative-pose prior terms—converts the separate pairwise extrinsic estimates into one globally consistent multi-camera calibration.","pith_inferences":["The same refinement pattern could extend to additional rigid sensors such as radar units on the same platform.","Downstream multi-view fusion tasks may benefit from the enforced geometric consistency even if they do not rerun the calibration.","The separation of learned initialization from geometric refinement suggests that similar two-stage pipelines could improve other sensor-calibration problems.","Real-world deployment may gain from running the refinement only when pairwise predictions show high mutual disagreement."],"forward_implications":["Per-camera translation error reaches 0.89 cm and rotation error reaches 0.038 on KITTI data.","Translation error drops from 108.6 cm to 3.1 cm on the Walkley dataset.","Inter-camera consistency improves beyond what independent pairwise processing achieves.","The joint step supplies robustness when individual camera predictions are less reliable."],"fun_headline_variants":["Bundle adjustment makes pairwise LiDAR estimates globally consistent","Learned pairwise init refined jointly for multi-camera LiDAR accuracy","Multi-frame refinement aligns independent camera-LiDAR extrinsic estimates","Two-stage calibration achieves consistent multi-camera LiDAR setup"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The bundle adjustment can find a globally consistent solution even when it starts from independent pairwise predictions that may contain errors.","fun_headline_variants_meta":{"raw":{"variants":["Bundle adjustment makes pairwise LiDAR estimates globally consistent","Learned pairwise init refined jointly for multi-camera LiDAR accuracy","Multi-frame refinement aligns independent camera-LiDAR extrinsic estimates","Two-stage calibration achieves consistent multi-camera LiDAR setup"]},"model":"grok-4.3","cost_usd":0.007941,"raw_usage":{"total_tokens":3600,"prompt_tokens":632,"num_sources_used":0,"completion_tokens":55,"cost_in_usd_ticks":79412000,"prompt_tokens_details":{"text_tokens":632,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2913,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":632,"tokens_out":55,"duration_ms":24380,"temperature":1.0,"reasoning_tokens":2913,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-28T22:30:37.321596+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A multi-camera dataset in which the final inter-camera relative poses after joint refinement deviate farther from ground truth than the original pairwise estimates.","supporting_citations":[],"review_version":1}