{"url":"/task/surface-normals-estimation","name":"Surface Normals Estimation","slug":"surface-normals-estimation","description_markdown":"Surface normal estimation deals with the task of predicting the surface orientation of the objects present inside a scene. Refer to [Designing Deep Networks for Surface Normal Estimation (Wang et al.)](https://www.cs.cmu.edu/~xiaolonw/papers/deep3d.pdf) to get a good overview of several design choices that led to the development of a CNN-based surface normal estimator.","categories":[{"name":"Computer Vision","url":"/area/computer-vision"}],"source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","slug_source":"archive_url"},"counts":{"papers_tagged":39,"papers_with_code":33,"benchmarks":8,"benchmark_tables_in_archive":8,"benchmark_tables_shown":8,"benchmark_tables_withheld_as_spam":0,"benchmark_definition":"a leaderboard table with at least one row; benchmark_tables_shown also counts the zero-row tables; benchmark_tables_in_archive adds the tables withheld as spam","datasets":12,"subtasks":0,"parent_tasks":0},"benchmarks":[{"leaderboard":"/sota/surface-normals-estimation-on-pcpnet","slug":"surface-normals-estimation-on-pcpnet","dataset":"PCPNet","dataset_url":null,"rows_in_archive":8,"metrics":["RMSE "],"first_row_in_archive_order":{"model":"MSECNet","paper_title":"MSECNet: Accurate and Robust Normal Estimation for 3D Point Clouds by Multi-Scale Edge Conditioning","paper_url":"/paper/msecnet-accurate-and-robust-normal-estimation","paper_date":"2023-08-04","arxiv_id":"2308.02237","code_links":[{"title":"martianxiu/MSECNet","url":"https://github.com/martianxiu/MSECNet"}],"syntology":null}},{"leaderboard":"/sota/surface-normals-estimation-on-stanford-orb","slug":"surface-normals-estimation-on-stanford-orb","dataset":"Stanford-ORB","dataset_url":"/dataset/stanford-orb","rows_in_archive":7,"metrics":["Cosine Distance"],"first_row_in_archive_order":{"model":"NVDiffRecMC","paper_title":"Shape, Light, and Material Decomposition from Images using Monte Carlo Rendering and Denoising","paper_url":"/paper/shape-light-material-decomposition-from","paper_date":"2022-06-07","arxiv_id":"2206.03380","code_links":[{"title":"NVlabs/nvdiffrecmc","url":"https://github.com/NVlabs/nvdiffrecmc"}],"syntology":null}},{"leaderboard":"/sota/surface-normals-estimation-on-nyu-depth-v2-1","slug":"surface-normals-estimation-on-nyu-depth-v2-1","dataset":"NYU Depth v2","dataset_url":"/dataset/nyuv2","rows_in_archive":6,"metrics":["% < 11.25","% < 22.5","% < 30","Mean Angle Error","RMSE"],"first_row_in_archive_order":{"model":"Metric3Dv2(L, FT)","paper_title":"Metric3Dv2: A Versatile Monocular Geometric Foundation Model for Zero-shot Metric Depth and Surface Normal Estimation","paper_url":"/paper/metric3d-v2-a-versatile-monocular-geometric-1","paper_date":"2024-03-22","arxiv_id":"2404.15506","code_links":[{"title":"yvanyin/metric3d","url":"https://github.com/yvanyin/metric3d"}],"syntology":{"n":3,"n_ran":3,"n_unverified":0,"n_pointer_only":0}}},{"leaderboard":"/sota/surface-normals-estimation-on-scannetv2","slug":"surface-normals-estimation-on-scannetv2","dataset":"ScanNetV2","dataset_url":"/dataset/scannet","rows_in_archive":3,"metrics":["% < 11.25","% < 22.5","% < 30","Mean Angle Error"],"first_row_in_archive_order":{"model":"Metric3Dv2 (g2, In-domain)","paper_title":"Metric3Dv2: A Versatile Monocular Geometric Foundation Model for Zero-shot Metric Depth and Surface Normal Estimation","paper_url":"/paper/metric3d-v2-a-versatile-monocular-geometric-1","paper_date":"2024-03-22","arxiv_id":"2404.15506","code_links":[{"title":"yvanyin/metric3d","url":"https://github.com/yvanyin/metric3d"}],"syntology":{"n":3,"n_ran":3,"n_unverified":0,"n_pointer_only":0}}},{"leaderboard":"/sota/surface-normals-estimation-on-ibims-1","slug":"surface-normals-estimation-on-ibims-1","dataset":"IBims-1","dataset_url":"/dataset/ibims-1","rows_in_archive":2,"metrics":["% < 11.25","% < 22.5","% < 30","Mean"],"first_row_in_archive_order":{"model":"Marigold + E2E FT(zero-shot)","paper_title":"Fine-Tuning Image-Conditional Diffusion Models is Easier than You Think","paper_url":"/paper/fine-tuning-image-conditional-diffusion","paper_date":"2024-09-17","arxiv_id":"2409.11355","code_links":[{"title":"VisualComputingInstitute/diffusion-e2e-ft","url":"https://github.com/VisualComputingInstitute/diffusion-e2e-ft"}],"syntology":{"n":17,"n_ran":16,"n_unverified":1,"n_pointer_only":17}}},{"leaderboard":"/sota/surface-normals-estimation-on-nyu-depth-v2","slug":"surface-normals-estimation-on-nyu-depth-v2","dataset":"NYU-Depth V2 Surface Normals","dataset_url":"/dataset/nyuv2","rows_in_archive":1,"metrics":["RMSE"],"first_row_in_archive_order":{"model":"DSN","paper_title":"On Deep Learning Techniques to Boost Monocular Depth Estimation for Autonomous Navigation","paper_url":"/paper/on-deep-learning-techniques-to-boost","paper_date":"2020-10-13","arxiv_id":"2010.06626","code_links":[],"syntology":null}},{"leaderboard":"/sota/surface-normals-estimation-on-pascal-context","slug":"surface-normals-estimation-on-pascal-context","dataset":"PASCAL Context","dataset_url":"/dataset/pascal-context","rows_in_archive":1,"metrics":["Mean Angle Error"],"first_row_in_archive_order":{"model":"InvPT","paper_title":"InvPT: Inverted Pyramid Multi-task Transformer for Dense Scene Understanding","paper_url":"/paper/inverted-pyramid-multi-task-transformer-for","paper_date":"2022-03-15","arxiv_id":"2203.07997","code_links":[{"title":"prismformore/InvPT","url":"https://github.com/prismformore/InvPT"}],"syntology":{"n":4,"n_ran":2,"n_unverified":2,"n_pointer_only":0}}},{"leaderboard":"/sota/surface-normals-estimation-on-taskonomy","slug":"surface-normals-estimation-on-taskonomy","dataset":"Taskonomy","dataset_url":"/dataset/taskonomy","rows_in_archive":1,"metrics":["L1 error"],"first_row_in_archive_order":{"model":"X-TC (Cross-Task Consistency)","paper_title":"Robust Learning Through Cross-Task Consistency","paper_url":"/paper/robust-learning-through-cross-task","paper_date":"2020-06-01","arxiv_id":null,"code_links":[{"title":"EPFL-VILAB/XTConsistency","url":"https://github.com/EPFL-VILAB/XTConsistency"}],"syntology":null}}],"datasets":[{"url":"/dataset/scannet","name":"ScanNet","full_name":"","num_papers_in_archive":1595},{"url":"/dataset/nyuv2","name":"NYUv2","full_name":"NYU-Depth V2","num_papers_in_archive":986},{"url":"/dataset/pascal-context","name":"PASCAL Context","full_name":"","num_papers_in_archive":323},{"url":"/dataset/taskonomy","name":"Taskonomy","full_name":"","num_papers_in_archive":147},{"url":"/dataset/ibims-1","name":"IBims-1","full_name":"Independent benchmark images and matched scans v1","num_papers_in_archive":34},{"url":"/dataset/grit","name":"GRIT","full_name":"General Robust Image Task Benchmark","num_papers_in_archive":16},{"url":"/dataset/stanford-orb","name":"Stanford-ORB","full_name":"","num_papers_in_archive":16},{"url":"/dataset/3d-ken-burns-dataset","name":"3D Ken Burns Dataset","full_name":"","num_papers_in_archive":4},{"url":"/dataset/pano3d","name":"Pano3D","full_name":"","num_papers_in_archive":2},{"url":"/dataset/supercaustics","name":"SuperCaustics","full_name":"","num_papers_in_archive":2},{"url":"/dataset/synfoot","name":"SynFoot","full_name":"","num_papers_in_archive":1},{"url":"/dataset/transproteus","name":"TransProteus","full_name":"","num_papers_in_archive":1}],"subtasks":[],"parent_tasks":[],"papers":{"order":"repositories listed in the archive (desc), then date (desc); the archive holds no stars","population":"papers tagged with this task that list at least one repository in the archive","shown":30,"of":33,"tagged_in_all":39,"items":[{"url":"/paper/blenderproc","title":"BlenderProc","date":"2019-10-25","arxiv_id":"1911.01911","repositories_listed":4,"syntology":null},{"url":"/paper/real-time-joint-semantic-segmentation-and","title":"Real-Time Joint Semantic Segmentation and Depth Estimation Using Asymmetric Annotations","date":"2018-09-13","arxiv_id":"1809.04766","repositories_listed":4,"syntology":{"n":6,"n_ran":6,"n_unverified":0,"n_pointer_only":6}},{"url":"/paper/idisc-internal-discretization-for-monocular","title":"iDisc: Internal Discretization for Monocular Depth Estimation","date":"2023-04-13","arxiv_id":"2304.06334","repositories_listed":2,"syntology":{"n":6,"n_ran":5,"n_unverified":1,"n_pointer_only":6}},{"url":"/paper/extracting-triangular-3d-models-materials-and","title":"Extracting Triangular 3D Models, Materials, and Lighting From Images","date":"2021-11-24","arxiv_id":"2111.12503","repositories_listed":2,"syntology":null},{"url":"/paper/360o-surface-regression-with-a-hyper-sphere","title":"$360^o$ Surface Regression with a Hyper-Sphere Loss","date":"2019-09-16","arxiv_id":"1909.07043","repositories_listed":2,"syntology":null},{"url":"/paper/differentiable-iterative-surface-normal","title":"Deep Iterative Surface Normal Estimation","date":"2019-04-15","arxiv_id":"1904.07172","repositories_listed":2,"syntology":null},{"url":"/paper/spherical-regression-learning-viewpoints","title":"Spherical Regression: Learning Viewpoints, Surface Normals and 3D Rotations on n-Spheres","date":"2019-04-10","arxiv_id":"1904.05404","repositories_listed":2,"syntology":{"n":3,"n_ran":3,"n_unverified":0,"n_pointer_only":3}},{"url":"/paper/fine-tuning-image-conditional-diffusion","title":"Fine-Tuning Image-Conditional Diffusion Models is Easier than You Think","date":"2024-09-17","arxiv_id":"2409.11355","repositories_listed":1,"syntology":{"n":17,"n_ran":16,"n_unverified":1,"n_pointer_only":17}},{"url":"/paper/polymax-general-dense-prediction-with-mask","title":"PolyMaX: General Dense Prediction with Mask Transformer","date":"2023-11-09","arxiv_id":"2311.05770","repositories_listed":1,"syntology":{"n":3,"n_ran":3,"n_unverified":0,"n_pointer_only":0}},{"url":"/paper/stanford-orb-a-real-world-3d-object-inverse","title":"Stanford-ORB: A Real-World 3D Object Inverse Rendering Benchmark","date":"2023-10-24","arxiv_id":"2310.16044","repositories_listed":1,"syntology":{"n":1,"n_ran":1,"n_unverified":0,"n_pointer_only":1}},{"url":"/paper/msecnet-accurate-and-robust-normal-estimation","title":"MSECNet: Accurate and Robust Normal Estimation for 3D Point Clouds by Multi-Scale Edge Conditioning","date":"2023-08-04","arxiv_id":null,"repositories_listed":1,"syntology":null},{"url":"/paper/mimic-masked-image-modeling-with-image","title":"MIMIC: Masked Image Modeling with Image Correspondences","date":"2023-06-27","arxiv_id":"2306.15128","repositories_listed":1,"syntology":null},{"url":"/paper/nefii-inverse-rendering-for-reflectance","title":"NeFII: Inverse Rendering for Reflectance Decomposition with Near-Field Indirect Illumination","date":"2023-03-29","arxiv_id":"2303.16617","repositories_listed":1,"syntology":{"n":11,"n_ran":1,"n_unverified":10,"n_pointer_only":0}},{"url":"/paper/neaf-learning-neural-angle-fields-for-point","title":"NeAF: Learning Neural Angle Fields for Point Normal Estimation","date":"2022-11-30","arxiv_id":"2211.16869","repositories_listed":1,"syntology":{"n":6,"n_ran":3,"n_unverified":3,"n_pointer_only":6}},{"url":"/paper/hsurf-net-normal-estimation-for-3d-point","title":"HSurf-Net: Normal Estimation for 3D Point Clouds by Learning Hyper Surfaces","date":"2022-10-13","arxiv_id":"2210.07158","repositories_listed":1,"syntology":null},{"url":"/paper/graphfit-learning-multi-scale-graph","title":"GraphFit: Learning Multi-scale Graph-Convolutional Representation for Point Cloud Normal Estimation","date":"2022-07-23","arxiv_id":"2207.11484","repositories_listed":1,"syntology":{"n":1,"n_ran":1,"n_unverified":0,"n_pointer_only":0}},{"url":"/paper/shape-light-material-decomposition-from","title":"Shape, Light, and Material Decomposition from Images using Monte Carlo Rendering and Denoising","date":"2022-06-07","arxiv_id":"2206.03380","repositories_listed":1,"syntology":null},{"url":"/paper/grit-general-robust-image-task-benchmark","title":"GRIT: General Robust Image Task Benchmark","date":"2022-04-28","arxiv_id":"2204.13653","repositories_listed":1,"syntology":{"n":2,"n_ran":2,"n_unverified":0,"n_pointer_only":0}},{"url":"/paper/inverted-pyramid-multi-task-transformer-for","title":"InvPT: Inverted Pyramid Multi-task Transformer for Dense Scene Understanding","date":"2022-03-15","arxiv_id":"2203.07997","repositories_listed":1,"syntology":{"n":4,"n_ran":2,"n_unverified":2,"n_pointer_only":0}},{"url":"/paper/estimating-and-exploiting-the-aleatoric","title":"Estimating and Exploiting the Aleatoric Uncertainty in Surface Normal Estimation","date":"2021-09-20","arxiv_id":"2109.09881","repositories_listed":1,"syntology":{"n":6,"n_ran":6,"n_unverified":0,"n_pointer_only":0}},{"url":"/paper/adafit-rethinking-learning-based-normal","title":"AdaFit: Rethinking Learning-based Normal Estimation on Point Clouds","date":"2021-08-12","arxiv_id":"2108.05836","repositories_listed":1,"syntology":{"n":17,"n_ran":9,"n_unverified":8,"n_pointer_only":17}},{"url":"/paper/nerfactor-neural-factorization-of-shape-and","title":"NeRFactor: Neural Factorization of Shape and Reflectance Under an Unknown Illumination","date":"2021-06-03","arxiv_id":"2106.01970","repositories_listed":1,"syntology":{"n":5,"n_ran":0,"n_unverified":5,"n_pointer_only":0}},{"url":"/paper/nerd-neural-reflectance-decomposition-from","title":"NeRD: Neural Reflectance Decomposition from Image Collections","date":"2020-12-07","arxiv_id":"2012.03918","repositories_listed":1,"syntology":null},{"url":"/paper/how-well-do-self-supervised-models-transfer","title":"How Well Do Self-Supervised Models Transfer?","date":"2020-11-26","arxiv_id":"2011.13377","repositories_listed":1,"syntology":{"n":5,"n_ran":4,"n_unverified":1,"n_pointer_only":0}},{"url":"/paper/ai-playground-unreal-engine-based-data","title":"AI Playground: Unreal Engine-based Data Ablation Tool for Deep Learning","date":"2020-07-13","arxiv_id":"2007.06153","repositories_listed":1,"syntology":null},{"url":"/paper/robust-learning-through-cross-task-1","title":"Robust Learning Through Cross-Task Consistency","date":"2020-06-07","arxiv_id":"2006.04096","repositories_listed":1,"syntology":{"n":1,"n_ran":1,"n_unverified":0,"n_pointer_only":1}},{"url":"/paper/sharingan-combining-synthetic-and-real-data-1","title":"SharinGAN: Combining Synthetic and Real Data for Unsupervised Geometry Estimation","date":"2020-06-07","arxiv_id":"2006.04026","repositories_listed":1,"syntology":null},{"url":"/paper/robust-learning-through-cross-task","title":"Robust Learning Through Cross-Task Consistency","date":"2020-06-01","arxiv_id":null,"repositories_listed":1,"syntology":null},{"url":"/paper/deepfit-3d-surface-fitting-via-neural-network","title":"DeepFit: 3D Surface Fitting via Neural Network Weighted Least Squares","date":"2020-03-23","arxiv_id":"2003.10826","repositories_listed":1,"syntology":null},{"url":"/paper/cleargrasp-3d-shape-estimation-of-transparent","title":"ClearGrasp: 3D Shape Estimation of Transparent Objects for Manipulation","date":"2019-10-06","arxiv_id":"1910.02550","repositories_listed":1,"syntology":{"n":8,"n_ran":0,"n_unverified":8,"n_pointer_only":0}}],"syntology_records":17,"syntology_note":"a paper without a record is not a recorded non-run: it may lack an arXiv id or simply be absent from the graph layer"},"description_links":{"kept":0,"unwrapped_to_text":0,"bare_urls_linked":0,"relative_images_dropped":0,"rule":"internal links are kept only when the target slug exists in the catalog"},"syntology":{"read_at":"2026-09-24T18:15:14+00:00","claim":"Per-sample execution status on synthesized fixtures ('ran N of M samples'); not a correctness claim and not a ranking signal.","status_vocabulary":{"ran_honours":"ran, honoured the contract we drafted","ran_violates":"ran, violated the contract we drafted","ran_draft_wrong":"ran; our contract draft was wrong, not the code","ran_fixture":"ran; our fixture could not drive it","ran":"ran on a synthesized input","unverified":"unverified (harvested, no recorded run)"}},"not_shown":{"libraries":"the archive has no per-task library table","trend_sparklines":"the Trend column of the benchmarks table was a rendered image; it is not in the archive","social_and_latest_sorts":"stars and social signals are not in the archive"}}