Papers › Floors are Flat: Leveraging Semantics for Real-Time Surface Normal Prediction

Floors are Flat: Leveraging Semantics for Real-Time Surface Normal Prediction

16 Jun 2019arXiv:1906.06792archive 2025-07-28

Steven Hickson, Karthik Raveendran, Alireza Fathi, Kevin Murphy, Irfan Essa

We propose 4 insights that help to significantly improve the performance of deep learning models that predict surface normals and semantic labels from a single RGB image. These insights are: (1) denoise the "ground truth" surface normals in the training set to ensure consistency with the semantic labels; (2) concurrently train on a mix of real and synthetic data, instead of pretraining on synthetic and finetuning on real; (3) jointly predict normals and semantics using a shared model, but only backpropagate errors on pixels that have valid training labels; (4) slim down the model and use grayscale instead of color inputs. Despite the simplicity of these steps, we demonstrate consistently improved results on several datasets, using a model that runs at 12 fps on a standard mobile phone.

PaperPDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

StevenHickson/CreateNormals officialmentioned on GitHub report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Semantic SegmentationSurface Normals Estimation

1 archive task tag without a task page not shown.

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Semantic Segmentation ScanNetV2 Floors are Flat Pixel Accuracy 65.6 #12 of 12 Archive leaderboard report
Surface Normals Estimation NYU Depth v2 Floors are Flat % < 11.25 59.5 #6 of 6 Archive leaderboard report
Surface Normals Estimation NYU Depth v2 Floors are Flat % < 22.5 72.2 #6 of 6 Archive leaderboard report
Surface Normals Estimation NYU Depth v2 Floors are Flat % < 30 77.3 #6 of 6 Archive leaderboard report
Surface Normals Estimation NYU Depth v2 Floors are Flat Mean Angle Error 19.7 #6 of 6 Archive leaderboard report
Surface Normals Estimation NYU Depth v2 Floors are Flat RMSE 19.3 #6 of 6 Archive leaderboard report
Surface Normals Estimation ScanNetV2 Floors are Flat % < 11.25 50.9 #3 of 3 Archive leaderboard report
Surface Normals Estimation ScanNetV2 Floors are Flat % < 22.5 65.2 #3 of 3 Archive leaderboard report
Surface Normals Estimation ScanNetV2 Floors are Flat % < 30 70 #3 of 3 Archive leaderboard report
Surface Normals Estimation ScanNetV2 Floors are Flat Mean Angle Error 28 #3 of 3 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections