Research map / Label-efficient and robust learning / Knowledge distillation
Monitored Distillation for Positive Congruent Depth Completion
Tian Yu Liu, Parth Agrawal, Allison Chen, Byung-Woo Hong, Alex Wong
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add sparse depth error weight beta to teacher selection (ablation) | Monitored Distillation · VOID (supervised ensemble) | MAE: 32.86 -> 30.88 | Improved |
| add sparse depth error weight beta to teacher selection (ablation) | Monitored Distillation · VOID (unsupervised ensemble) | MAE: 38.78 -> 36.42 | Improved |
| add sparse depth error weight beta to teacher selection (ablation) | Monitored Distillation · VOID (heterogeneous ensemble) | MAE: 32.09 -> 29.67 | Improved |
| add unsupervised loss to mean distillation (ablation) | Mean ensemble distillation · VOID (heterogeneous ensemble) | MAE: 44.53 -> 35.79 | Improved |
| add unsupervised loss to mean distillation (ablation) | Mean ensemble distillation · VOID (supervised ensemble) | MAE: 34.27 -> 34.04 | No clear effect |
| replace mean ensembling with monitored distillation | Mean ensemble distillation · VOID (heterogeneous ensemble) | MAE: 35.79 -> 29.67 | Improved |
| add weaker teachers to a naive mean ensemble (supervised to heterogeneous) | Mean ensemble distillation · VOID | MAE: 34.04 -> 35.79 | Hurt |
| add ScaffNet to the FusionNet and KBNet ensemble (ablation) | Monitored Distillation · VOID | MAE: 40.10 -> 36.42 | Improved |
| train teachers on NYUv2 and SceneNet instead of VOID (ablation) | Monitored Distillation · VOID | MAE: 36.42 -> 46.66 | Hurt |
| distill from monitored blind ensemble instead of single best teacher NLSPN | Distill NLSPN · KITTI validation | MAE: 221.077 -> 218.222 | Improved |
| distill from single teacher instead of mean of ensemble (ablation) | Mean ensemble distillation · KITTI validation | MAE: 232.481 -> 221.077 | Improved |
| reduce sparse depth density from 0.50% to 0.05% | Monitored Distillation · VOID | MAE: 29.666 -> 104.966 | Hurt |
Closest papers by meaning
- Distilling Monocular Foundation Model for Fine-grained Depth Completion (CVPR 2025 · similarity 0.88)
- HFD-Teacher: High-Frequency Depth Distillation from Depth Foundation Models for Enhanced Depth Completion (ICCV 2025 · similarity 0.87)
- Test-Time Adaptation for Depth Completion (CVPR 2024 · similarity 0.86)
- Self-Distilled Depth Refinement with Noisy Poisson Fusion (NeurIPS 2024 · similarity 0.86)
- Zero-shot Depth Completion via Test-time Alignment with Affine-invariant Depth Prior (AAAI 2025 · similarity 0.85)
- Flexible Depth Completion for Sparse and Varying Point Densities (CVPR 2024 · similarity 0.85)
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.
