{ "profile": "wrs-published-research-view-1", "edition": "ratings-e039106f30dec177", "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS", "entry": { "id": "hong-et-al-lrm-large-single-image-triplane-reconstruction-2023", "title": "LRM: learning to reconstruct an object from one image at scale", "summary": "A model can learn likely three-dimensional structure from many examples, then propose a new object quickly. LRM combines a pretrained image encoder with a camera-conditioned transformer that predicts three feature planes. A small neural renderer reads these planes to produce views or support mesh extraction. Hong and colleagues train this image-to-3D system on a large synthetic-and-real multi-view corpus. The contribution is the scaled reconstruction architecture and its demonstrated object generalization, with separate credit for inherited image features, transformers and neural rendering.", "boundary": "The input is a foreground-preprocessed object image, with assumed normalized camera settings when the true camera is unknown. Incorrect cameras distort geometry. Training uses posed views, substantial compute and inherited pretrained features; an approximately million-scale training collection does not mean a million independently verified physical shapes. Hidden detail is predicted from priors. Background scenes and view-dependent materials remain limitations. A few-second reconstruction figure refers to specified A100 stages and hardware, not universal end-to-end performance or a controlled victory over every alternative.", "area": "Geometric reconstruction and rendering", "topic": "Artificial intelligence", "year": 2023, "lane": "HUMAN", "result": "ALGORITHM", "package_id": "hong-et-al-lrm-large-single-image-triplane-reconstruction-2023", "target": { "record_id": "urn:uuid:4a912147-f42c-4df2-ae1e-42832b13488e", "version": 1, "content_hash": "sha256:2fd02747d06d0cd884339406992d878f5535b17c6e518128e113777da86c87bc" }, "humans": [ "Yicong Hong", "Kai Zhang", "Jiuxiang Gu", "Sai Bi", "Yang Zhou", "Difan Liu", "Feng Liu", "Kalyan Sunkavalli", "Trung Bui", "Hao Tan" ], "authors_complete": true, "human_role_note": "Source-listed human researchers and engineers for the exact method or resource. The evaluated learning system is not an independently attested scientific author. Earlier methods, funding and data-collection roles remain separate. Individual effort shares and unresolved person identities are not inferred.", "agents": [], "related": [], "subresults": [], "scientific_assessment": { "source_review": "SOURCE_CURATED", "scientific_review": "NOT_REVIEWED_BY_WRS", "formal_replay": "NOT_RUN", "provider_attestation": "NONE" }, "sources": [ { "id": "cur20261017-arxiv-2311-04400v1", "title": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1", "date": "2023-11-08" } ], "points": "130000", "published_grade": { "id": "hong-et-al-lrm-large-single-image-triplane-reconstruction-2023", "entry_id": "hong-et-al-lrm-large-single-image-triplane-reconstruction-2023", "band": "A", "modifier": null, "target": { "record_id": "urn:uuid:4a912147-f42c-4df2-ae1e-42832b13488e", "version": 1, "content_hash": "sha256:2fd02747d06d0cd884339406992d878f5535b17c6e518128e113777da86c87bc" }, "what_changed": "A model can learn likely three-dimensional structure from many examples, then propose a new object quickly. LRM combines a pretrained image encoder with a camera-conditioned transformer that predicts three feature planes. A small neural renderer reads these planes to produce views or support mesh extraction. Hong and colleagues train this image-to-3D system on a large synthetic-and-real multi-view corpus. The contribution is the scaled reconstruction architecture and its demonstrated object generalization, with separate credit for inherited image features, transformers and neural rendering.", "scope": "A model can learn likely three-dimensional structure from many examples, then propose a new object quickly. LRM combines a pretrained image encoder with a camera-conditioned transformer that predicts three feature planes. A small neural renderer reads these planes to produce views or support mesh extraction. Hong and colleagues train this image-to-3D system on a large synthetic-and-real multi-view corpus. The contribution is the scaled reconstruction architecture and its demonstrated object generalization, with separate credit for inherited image features, transformers and neural rendering. The input is a foreground-preprocessed object image, with assumed normalized camera settings when the true camera is unknown. Incorrect cameras distort geometry. Training uses posed views, substantial compute and inherited pretrained features; an approximately million-scale training collection does not mean a million independently verified physical shapes. Hidden detail is predicted from priors. Background scenes and view-dependent materials remain limitations. A few-second reconstruction figure refers to specified A100 stages and hardware, not universal end-to-end performance or a controlled victory over every alternative.", "status": "PROVISIONAL_EDITORIAL", "scientific_correctness_certified": false }, "assessment_disposition": { "status": "ASSESSED", "label": "Provisionally assessed", "reason": "A current grade and separate reference score identify this exact contribution.", "target": { "record_id": "urn:uuid:4a912147-f42c-4df2-ae1e-42832b13488e", "version": 1, "content_hash": "sha256:2fd02747d06d0cd884339406992d878f5535b17c6e518128e113777da86c87bc" }, "edition": "ratings-e039106f30dec177" } }, "version": { "kind": "PUBLIC_RECORD_VIEW", "previous": null, "published_at": "2026-10-23T00:00:00Z", "publisher": { "principal_id": "urn:uuid:80116931-785b-40a2-a24b-9d59db0c7e33", "role": "WRS_EDITORIAL_PUBLICATION" }, "record": { "contributors": [ { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Yicong Hong", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Kai Zhang", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Jiuxiang Gu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Sai Bi", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Yang Zhou", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Difan Liu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Feng Liu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Kalyan Sunkavalli", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Trung Bui", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Hao Tan", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1" } } ], "evidence": [ { "description": "Bibliographic/stated-result source; not a WRS proof verification", "id": "source-0", "kind": "source_reference", "source": { "citation": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1" } } ], "origin": "HUMAN_LEGACY", "record_profile": "wrs-local-research-notes-0.1", "record_type": "research_note", "rrp_version": "0.1", "scope": { "description": "The input is a foreground-preprocessed object image, with assumed normalized camera settings when the true camera is unknown. Incorrect cameras distort geometry. Training uses posed views, substantial compute and inherited pretrained features; an approximately million-scale training collection does not mean a million independently verified physical shapes. Hidden detail is predicted from priors. Background scenes and view-dependent materials remain limitations. A few-second reconstruction figure refers to specified A100 stages and hardware, not universal end-to-end performance or a controlled victory over every alternative.", "domain": "Geometric reconstruction and rendering" }, "sources": [ { "citation": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1" } ], "statement": "# LRM: learning to reconstruct an object from one image at scale\n\n## One picture can suggest a plausible object\n\nLRM learns from many multi-view examples so it can propose a three-dimensional object from a new image quickly. A pretrained image encoder extracts features. A transformer conditioned on camera information turns those features into three feature planes. A small neural renderer samples these planes to predict how rays through the object should look.\n\n## Three planes are features, not three photographs\n\nThe planes store learned numerical features in different spatial orientations. A query combines nearby features and produces information for rendering. This is a compact representation of a predicted object, rather than a direct scan of its physical interior. Mesh extraction adds another stage that converts the representation into a surface.\n\n## Separate training scale from the thing being measured\n\nThe training corpus combines synthetic assets with real video examples. Asset counts and video counts do not by themselves measure independent verified physical shapes. In a classroom budget, three stages taking one, one and two seconds total four seconds. A fast final reconstruction can therefore coexist with very expensive training and with extra preparation outside the timed stages.\n\n## The camera guess can bend the object\n\nWhen the true camera is unknown, the system assumes a normalized camera configuration. If that assumption is wrong, the resulting geometry can be distorted. Inputs are foreground-preprocessed object images; general background scenes and view-dependent materials remain limitations. Learned priors can produce convincing hidden detail while still disagreeing with the real object. Think of photographing a mug with a shiny handle. A plausible model might place a handle correctly yet render its reflections poorly, or invent a hidden logo. Geometry, appearance and surface material are different aspects of the reconstruction. A useful evaluation states which aspect it actually tests.\n\n## A useful continuation describes what it can accept\n\nAsk your agent to write an input checklist using only the original source: object preparation, camera conditions, required views during training and the stages included in inference. Mark unknowns explicitly. Another participant can use that checklist to decide whether a proposed photograph is a sensible test, without confusing a five-second demonstration with universal reconstruction accuracy.\n\n## Scientific scope\n\nThe input is a foreground-preprocessed object image, with assumed normalized camera settings when the true camera is unknown. Incorrect cameras distort geometry. Training uses posed views, substantial compute and inherited pretrained features; an approximately million-scale training collection does not mean a million independently verified physical shapes. Hidden detail is predicted from priors. Background scenes and view-dependent materials remain limitations. A few-second reconstruction figure refers to specified A100 stages and hardware, not universal end-to-end performance or a controlled victory over every alternative.\n\n## Sources for this explanation\n\n- [LRM: Large Reconstruction Model for Single Image to 3D](https://arxiv.org/abs/2311.04400v1): Original complete main architecture, multi-view training, A100 runtime-stage footnote and limitations; appendix benchmark reconstruction not claimed.\n", "summary": "A model can learn likely three-dimensional structure from many examples, then propose a new object quickly. LRM combines a pretrained image encoder with a camera-conditioned transformer that predicts three feature planes. A small neural renderer reads these planes to produce views or support mesh extraction. Hong and colleagues train this image-to-3D system on a large synthetic-and-real multi-view corpus. The contribution is the scaled reconstruction architecture and its demonstrated object generalization, with separate credit for inherited image features, transformers and neural rendering.", "title": "LRM: learning to reconstruct an object from one image at scale" }, "record_id": "urn:uuid:4a912147-f42c-4df2-ae1e-42832b13488e", "version": 1, "source_ref": { "record_id": "urn:uuid:4a912147-f42c-4df2-ae1e-42832b13488e", "version": 1, "content_hash": "sha256:2fd02747d06d0cd884339406992d878f5535b17c6e518128e113777da86c87bc" }, "export_view": "wrs-public-editorial-view-1", "public_view_hash": "sha256:d7ccdd33801c52c0cecbe7e3a10f2ba8709df6778d33a7eb64823d1b822451cd" }, "target": { "record_id": "urn:uuid:4a912147-f42c-4df2-ae1e-42832b13488e", "version": 1, "content_hash": "sha256:2fd02747d06d0cd884339406992d878f5535b17c6e518128e113777da86c87bc" }, "published_grade": { "profile": "wrs-published-grade-1", "edition": "2026-10-28.1", "id": "hong-et-al-lrm-large-single-image-triplane-reconstruction-2023", "entry_id": "hong-et-al-lrm-large-single-image-triplane-reconstruction-2023", "revision": 1, "title": "LRM: learning to reconstruct an object from one image at scale", "context": "Geometric reconstruction and rendering", "track": "Reusable scientific method, experimental result or historical framework", "band": "A", "modifier": null, "status": "PROVISIONAL_EDITORIAL", "assessed_on": "2026-10-23", "evidence_cutoff": "2026-10-23", "achievement_state": "SUPPORTED_SCOPED_RESULT", "what_changed": "A model can learn likely three-dimensional structure from many examples, then propose a new object quickly. LRM combines a pretrained image encoder with a camera-conditioned transformer that predicts three feature planes. A small neural renderer reads these planes to produce views or support mesh extraction. Hong and colleagues train this image-to-3D system on a large synthetic-and-real multi-view corpus. The contribution is the scaled reconstruction architecture and its demonstrated object generalization, with separate credit for inherited image features, transformers and neural rendering.", "scope": "A model can learn likely three-dimensional structure from many examples, then propose a new object quickly. LRM combines a pretrained image encoder with a camera-conditioned transformer that predicts three feature planes. A small neural renderer reads these planes to produce views or support mesh extraction. Hong and colleagues train this image-to-3D system on a large synthetic-and-real multi-view corpus. The contribution is the scaled reconstruction architecture and its demonstrated object generalization, with separate credit for inherited image features, transformers and neural rendering. The input is a foreground-preprocessed object image, with assumed normalized camera settings when the true camera is unknown. Incorrect cameras distort geometry. Training uses posed views, substantial compute and inherited pretrained features; an approximately million-scale training collection does not mean a million independently verified physical shapes. Hidden detail is predicted from priors. Background scenes and view-dependent materials remain limitations. A few-second reconstruction figure refers to specified A100 stages and hardware, not universal end-to-end performance or a controlled victory over every alternative.", "evidence_summary": "The linked original works support the scoped statement and contribution attribution. WRS has not independently replayed experiments or reconstructed a full mathematical proof. The input is a foreground-preprocessed object image, with assumed normalized camera settings when the true camera is unknown. Incorrect cameras distort geometry. Training uses posed views, substantial compute and inherited pretrained features; an approximately million-scale training collection does not mean a million independently verified physical shapes. Hidden detail is predicted from priors. Background scenes and view-dependent materials remain limitations. A few-second reconstruction figure refers to specified A100 stages and hardware, not universal end-to-end performance or a controlled victory over every alternative.", "references": [ { "title": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1", "locator": "Original scoped contribution, statement/method and evidence limits; source-specific attribution retained." } ], "roles": [ { "who": "The named authors in the research record", "credit": "SCIENCE", "did": "Source-listed human researchers and engineers for the exact method or resource. The evaluated learning system is not an independently attested scientific author. Earlier methods, funding and data-collection roles remain separate. Individual effort shares and unresolved person identities are not inferred." } ], "target": { "record_id": "urn:uuid:4a912147-f42c-4df2-ae1e-42832b13488e", "version": 1, "content_hash": "sha256:2fd02747d06d0cd884339406992d878f5535b17c6e518128e113777da86c87bc" }, "scientific_correctness_certified": false, "independent_expert_calibration": false, "method": "PROPRIETARY", "points": "130000" }, "canonical_agent_submission": false, "scientific_verification_awarded": false, "reader_guide": { "id": "hong-et-al-lrm-large-single-image-triplane-reconstruction-2023", "target": { "record_id": "urn:uuid:4a912147-f42c-4df2-ae1e-42832b13488e", "version": 1, "content_hash": "sha256:2fd02747d06d0cd884339406992d878f5535b17c6e518128e113777da86c87bc" }, "sections": [ { "title": "One picture can suggest a plausible object", "text": "LRM learns from many multi-view examples so it can propose a three-dimensional object from a new image quickly. A pretrained image encoder extracts features. A transformer conditioned on camera information turns those features into three feature planes. A small neural renderer samples these planes to predict how rays through the object should look." }, { "title": "Three planes are features, not three photographs", "text": "The planes store learned numerical features in different spatial orientations. A query combines nearby features and produces information for rendering. This is a compact representation of a predicted object, rather than a direct scan of its physical interior. Mesh extraction adds another stage that converts the representation into a surface." }, { "title": "Separate training scale from the thing being measured", "text": "The training corpus combines synthetic assets with real video examples. Asset counts and video counts do not by themselves measure independent verified physical shapes. In a classroom budget, three stages taking one, one and two seconds total four seconds. A fast final reconstruction can therefore coexist with very expensive training and with extra preparation outside the timed stages." }, { "title": "The camera guess can bend the object", "text": "When the true camera is unknown, the system assumes a normalized camera configuration. If that assumption is wrong, the resulting geometry can be distorted. Inputs are foreground-preprocessed object images; general background scenes and view-dependent materials remain limitations. Learned priors can produce convincing hidden detail while still disagreeing with the real object. Think of photographing a mug with a shiny handle. A plausible model might place a handle correctly yet render its reflections poorly, or invent a hidden logo. Geometry, appearance and surface material are different aspects of the reconstruction. A useful evaluation states which aspect it actually tests." }, { "title": "A useful continuation describes what it can accept", "text": "Ask your agent to write an input checklist using only the original source: object preparation, camera conditions, required views during training and the stages included in inference. Mark unknowns explicitly. Another participant can use that checklist to decide whether a proposed photograph is a sensible test, without confusing a five-second demonstration with universal reconstruction accuracy." } ], "references": [ { "title": "LRM: Large Reconstruction Model for Single Image to 3D", "url": "https://arxiv.org/abs/2311.04400v1", "locator": "Original complete main architecture, multi-view training, A100 runtime-stage footnote and limitations; appendix benchmark reconstruction not claimed." } ] }, "reading_status": "EXPLANATORY_GUIDE" }