AI Paper Atlas

Research map / 3D vision / Depth and stereo

Self-Supervised Deep Depth Denoising

Vladimiros Sterzentsenko, Leonidas Saroglou, Anargyros Chatzitofis, Spyridon Thermos, Nikolaos Zioulis, Alexandros Doumanoglou, Dimitrios Zarpalas, Petros Daras

ICCV 2019 · 3D vision › Depth and stereo · approach: Dense depth prediction and completion · 48 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add photometric multi-view supervision instead of plain autoencoder reconstruction (ablation)self-supervised denoiser · D415 test set (70 pairs, K2 reference)MAE (mm): 26.35 -> 25.96Improved
add photometric supervision instead of plain autoencoder (ablation)self-supervised denoiser · D415 test set (70 pairs, K2 reference)RMSE (mm): 59.92 -> 58.30Improved
add BerHu depth regularization to photometric loss (ablation)P-only · D415 test set (70 pairs, K2 reference)MAE (mm): 25.96 -> 26.43Hurt
add BerHu depth regularization to photometric loss (ablation)P-only · D415 test set (70 pairs, K2 reference)Normal angle within 30 deg (%): 53.11 -> 53.98Improved
add normal smoothness prior to photometric loss (no depth term) (ablation)P-only · D415 test set (70 pairs, K2 reference)MAE (mm): 25.96 -> 28.04Hurt
add normal prior on top of photometric plus depth loss (ablation)P+D · D415 test set (70 pairs, K2 reference)MAE (mm): 26.43 -> 25.11Improved
add normal prior on top of photometric plus depth loss (ablation)P+D · D415 test set (70 pairs, K2 reference)RMSE (mm): 58.31 -> 58.95Hurt
add depth loss to photometric plus normal prior (ablation)P+N · D415 test set (70 pairs, K2 reference)MAE (mm): 28.04 -> 25.11Improved
use proposed self-supervised model instead of DRR (ablation)DRR · InteriorNet (synthetic noise)MAE: 86.88 -> 33.44Improved
train DDRNet on our dataset with partial convolutions instead of traditionalDDRNet retrained · D415 test set (70 pairs, K2 reference)MAE (mm): 121.83 -> 75.68Improved
use our model instead of DDRNet-PC retrained on our dataDDRNet-PC · D415 test set (70 pairs, K2 reference)MAE (mm): 75.68 -> 25.11Improved
use our model instead of RGF for KinectFusionRGF · KinectFusion reconstruction (15 sequences)point-to-plane RMSE (mm): 32.58 -> 24.74Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy