{"id":"18b78897-6c8e-4221-8eef-1c78f4dea4ac","arxiv_id":"2606.17824","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A human-in-the-loop pipeline generates usable segmented 2D atlases from diverse 3D geometries by using greedy view selection, SAM 2 interactive segmentation, and UV back-projection, with recurring manual corrections needed for fine structures, cavities, and weak boundaries.","lead":"The paper describes a pipeline that picks a few 2D views of a 3D model, lets users segment them interactively with SAM 2 in a labeling tool, and maps the segments back to a single texture atlas via UV coordinates. Smart generalists might read it to see how human guidance can make AI tools practical for preparing 3D assets in games and XR without full manual 3D editing.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Back-projection of 2D SAM2 masks to UV atlas lacks quantitative validation of fidelity under occlusion and view selection.","rationale":"The reader’s weakest assumption matches the projection step exactly. The demonstration-only evaluation supplies no counter-evidence, so the UNVERDICTED / LOW verdict stands. No other internal inconsistency (e.g., in the set-cover formulation or SAM2 usage) rises to the same load-bearing level.","tokens_in":1648,"tokens_out":360,"duration_ms":17699,"concrete_test":"For each of the eight meshes, render the final atlas back onto the original views and compute per-pixel agreement (IoU and boundary error) against the interactively corrected 2D masks; also measure the fraction of UV texels receiving conflicting labels from multiple views. If mean IoU drops below 0.85 or >3 % of texels show label conflicts, the usability claim is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that masks from a greedy set-cover view selection, once back-projected, yield a usable unified atlas without material distortion, overlap conflicts, or loss of fine detail. The method description implies standard rasterization or ray-casting back-projection onto the existing UV map, yet the only evidence is a demonstration on eight cultural-heritage meshes. No metric (IoU, boundary F-score, or per-texel consistency across overlapping views) is reported for the projection step itself. For geometries containing cavities or thin structures, even modest view-selection gaps or depth-buffer errors would produce atlas defects that downstream material assignment cannot tolerate. This is the least-secured link because the pipeline’s novelty rests on the atlas being production-ready after minimal manual fixes.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper presents a human-in-the-loop pipeline for atlas-based segmentation of 3D assets. It selects a compact set of rendered views via greedy set cover over surface points, performs interactive 2D segmentation using SAM 2 in Label Studio, and back-projects the resulting masks onto the model's UV parameterization to produce a unified segmented atlas suitable for material assignment, style transfer, and semantic labeling. Feasibility is assessed via a qualitative demonstration on eight cultural heritage objects, which identifies recurring manual correction needs for fine structures, cavities, and weak appearance boundaries.","tokens_in":1794,"tokens_out":525,"duration_ms":25070,"significance":"If the back-projection step preserves segmentation fidelity without significant distortion or conflicts, the pipeline could offer a practical, controllable workflow for interactive 3D content creation in games, media, and XR. The explicit identification of common failure modes (cavities, thin structures) provides actionable guidance for refinement. The absence of quantitative validation, however, limits the strength of claims about usability and generalizability across geometries.","major_comments":[{"comment":"Evaluation section: The demonstration on eight objects reports no quantitative metrics for the back-projection step itself (e.g., no IoU, boundary F-score, per-texel consistency across overlapping views, or error rates under occlusion and depth-buffer artifacts). This directly weakens the central claim that the method yields 'usable' atlases after minimal manual fixes, as the skeptic correctly identifies this as the least-secured link in the pipeline.","section":"Evaluation"},{"comment":"Pipeline description (back-projection paragraph): The method implies standard rasterization or ray-casting onto the existing UV map, yet provides no details on conflict resolution for overlapping views, handling of cavities/thin structures, or view-selection gaps. Without these, the claim that the atlas supports downstream production tasks remains unverified for the geometries highlighted as problematic.","section":"Method"}],"minor_comments":[{"comment":"The coverage threshold parameter in the greedy set cover is mentioned but not specified (value, sensitivity, or per-model tuning), which affects reproducibility of the view selection.","section":null},{"comment":"Abstract uses 'SAM~2'; standardize notation to 'SAM 2' throughout for consistency.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a methods/demonstration paper rather than a theoretical contribution; its fit for a core CV journal depends on whether the editors value application pipelines. The lack of baselines or comparisons to prior 3D segmentation or atlas methods further weakens positioning."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the detailed and constructive review. We address each major comment below, clarifying the scope of our demonstration-based evaluation while committing to targeted revisions that strengthen the manuscript without altering its core positioning as a practical workflow description.","responses":[{"response":"We appreciate this observation. Our evaluation is deliberately qualitative and demonstration-based, centered on end-to-end usability for diverse cultural heritage geometries and the explicit identification of recurring correction needs (fine structures, cavities, weak boundaries). Quantitative metrics such as IoU or boundary F-score presuppose application-independent ground truth, which does not exist for these objects. In revision we will add a dedicated paragraph in the Evaluation section that (a) acknowledges this limitation, (b) reports view-consistency statistics (per-texel label agreement across overlapping projections) on the existing data, and (c) outlines how future users could compute task-specific metrics once ground truth is defined. This provides additional transparency without overstating the current evidence.","revision_made":"partial","referee_comment":"[Evaluation] Evaluation section: The demonstration on eight objects reports no quantitative metrics for the back-projection step itself (e.g., no IoU, boundary F-score, per-texel consistency across overlapping views, or error rates under occlusion and depth-buffer artifacts). This directly weakens the central claim that the method yields 'usable' atlases after minimal manual fixes, as the skeptic correctly identifies this as the least-secured link in the pipeline."},{"response":"We agree that the back-projection description is underspecified. The revised manuscript will expand the relevant paragraph to state: (1) conflict resolution uses a priority-weighted majority vote based on view normal alignment and coverage; (2) cavities and thin structures are automatically flagged when depth discontinuities exceed a threshold and are routed to the interactive correction stage, consistent with the failure modes already reported; (3) residual view-selection gaps are mitigated by permitting the user to request additional views within Label Studio. These clarifications will directly address how the pipeline remains viable for the geometries discussed.","revision_made":"yes","referee_comment":"[Method] Pipeline description (back-projection paragraph): The method implies standard rasterization or ray-casting onto the existing UV map, yet provides no details on conflict resolution for overlapping views, handling of cavities/thin structures, or view-selection gaps. Without these, the claim that the atlas supports downstream production tasks remains unverified for the geometries highlighted as problematic."}],"tokens_in":1372,"tokens_out":529,"duration_ms":23951,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The core contribution is a pipeline that picks a small set of views with greedy set cover, lets a user segment them interactively with SAM 2 in Label Studio, and back-projects the masks onto the existing UV parameterization. That combination is not in the prior work they cite, and the demonstration on eight cultural-heritage meshes shows it can produce atlases that are close enough for downstream material assignment after limited fixes.\n\nWhat stands out is the honest reporting of recurring trouble spots: fine structures, cavities, and weak boundaries. That kind of concrete feedback is useful for anyone building similar tools.\n\nThe evaluation is the clear limitation. Everything is qualitative. There are no IoU numbers, no boundary-error measures, no comparison against simpler projection baselines, and no test of how well overlapping views agree after back-projection. The stress-test concern holds: without those checks it is hard to know whether the atlas is actually production-ready or just acceptable for the eight demo cases. The free parameter (coverage threshold) is mentioned but not explored.\n\nThis paper is for practitioners who need a working workflow for game or XR asset segmentation rather than for theorists looking for new algorithms. A reader who cares about applied 3D content pipelines will find the description and the identified failure modes worth the time.\n\nIt deserves peer review. The problem is real, the method is straightforward and reproducible in principle, and the main gap (quantitative projection validation) is fixable with additional experiments rather than a fundamental flaw.","headline":"This is a practical engineering paper on a SAM-2-based human-in-the-loop pipeline for turning 3D meshes into segmented UV atlases, but the back-projection step has no quantitative validation.","tokens_in":2258,"tokens_out":386,"would_cite":false,"duration_ms":17243,"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":"A pipeline selects a few 2D views of a 3D model, lets users segment them interactively, and back-projects the masks to a single UV atlas for downstream editing.","keywords":["3D asset segmentation","human-in-the-loop","UV atlas","interactive segmentation","view selection","back-projection","cultural heritage","SAM2"],"falsifier":"Compare the final atlas against a manually painted ground-truth segmentation on the 3D surface for a model with cavities or thin features and measure the fraction of surface area that mismatches after projection.","tokens_in":2553,"feed_emoji":"🖼️","tokens_out":728,"duration_ms":25674,"temperature":0.7,"pith_summary":"The paper describes a method that turns 3D asset segmentation into a manageable 2D task by first picking a small set of rendered views that together cover the surface, then letting a user refine masks on those views, and finally mapping the results back onto the model's UV layout. This setup keeps human judgment in the loop for criteria that depend on the final use case, such as material or style decisions in games and XR. Evaluation on eight cultural heritage objects shows the resulting atlases are usable across varied shapes, while also identifying the kinds of regions that still need manual fixes. The approach matters because full 3D segmentation without user guidance often fails when boundaries are application-specific or visually weak.","feed_headline":"Few 2D views plus back-projection produce segmented 3D atlases","feed_subtitle":"Greedy coverage of surface points followed by SAM2 masking and UV transfer supports material and style tasks on varied models.","key_machinery":"Greedy set-cover view selection followed by back-projection of 2D masks onto the 3D model's UV parameterization.","core_discovery":"The method generates segmented 2D parameterized atlases from 3D models by first choosing a compact set of rendered views via greedy set cover on sampled surface points, then performing interactive segmentation on those views using SAM 2 and Label Studio, and finally back-projecting the masks onto the UV parameterization to yield a unified atlas suitable for tasks like material assignment and style transfer. Testing on eight cultural heritage objects confirms that usable atlases result for diverse geometries, with recurring needs for correction on fine structures, cavities, and weak appearance boundaries.","pith_inferences":["The method could shorten production time for non-heritage assets such as game props if the same view-coverage logic holds.","If surface sampling misses thin protrusions, the greedy view set may leave gaps that require extra manual masks beyond what the paper reports.","Combining the atlas output with real-time engines might allow artists to see live updates when they adjust a 2D mask."],"forward_implications":["Material assignment and style transfer can be performed region by region directly on the atlas.","Semantic labels produced on the atlas transfer to the 3D model for use in game or XR pipelines.","The same view-selection and projection steps can be reused for any 3D model that has a UV parameterization.","Recurring correction patterns on fine structures and cavities indicate where further automation would reduce user effort most."],"fun_headline_variants":["Greedy sampling selects views for SAM2-based 3D atlas creation","Human-in-the-loop SAM2 segmentation generates UV atlases from 3D","Compact 3D views segmented interactively yield unified asset atlases","UV backprojection turns SAM2 view masks into segmented 3D atlases"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"Back-projecting the 2D masks onto the UV layout transfers the segmentation without major distortion, overlaps, or loss of detail from the chosen views.","fun_headline_variants_meta":{"raw":{"variants":["Greedy sampling selects views for SAM2-based 3D atlas creation","Human-in-the-loop SAM2 segmentation generates UV atlases from 3D","Compact 3D views segmented interactively yield unified asset atlases","UV backprojection turns SAM2 view masks into segmented 3D atlases"]},"model":"grok-4.3","cost_usd":0.005231,"raw_usage":{"total_tokens":2523,"prompt_tokens":646,"num_sources_used":0,"completion_tokens":71,"cost_in_usd_ticks":52312000,"prompt_tokens_details":{"text_tokens":646,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1806,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":646,"tokens_out":71,"duration_ms":18557,"temperature":1.0,"reasoning_tokens":1806,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-27T01:53:40.884212+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Compare the final atlas against a manually painted ground-truth segmentation on the 3D surface for a model with cavities or thin features and measure the fraction of surface area that mismatches after projection.","supporting_citations":[],"review_version":1}