Research map / Architectures and efficiency / CNN design
AdaBins: Depth Estimation Using Adaptive Bins
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 per-image adaptive depth bins with hybrid regression (ablation) | EfficientNet B5 encoder-decoder · NYU-Depth-v2 test | δ1: 0.881 -> 0.903 | Helped |
| Add per-image adaptive depth bins with hybrid regression (ablation) | EfficientNet B5 encoder-decoder · NYU-Depth-v2 test | REL: 0.111 -> 0.103 | Helped |
| Replace standard regression with fixed uniform bins and hybrid regression (ablation) | EfficientNet B5 encoder-decoder · NYU-Depth-v2 test | δ1: 0.881 -> 0.892 | Helped |
| Replace per-image adaptive bins with fixed log-uniform bins (ablation) | AdaBins depth estimator · NYU-Depth-v2 test | δ1: 0.896 -> 0.903 | Hurt |
| Replace per-image adaptive bins with dataset-trained fixed bins (ablation) | AdaBins depth estimator · NYU-Depth-v2 test | δ1: 0.893 -> 0.903 | Hurt |
| Remove the bin-center Chamfer loss (ablation) | AdaBins · NYU-Depth-v2 test | REL: 0.103 -> 0.106 | Hurt |
| Replace SI+Bins loss with L1/SSIM loss (ablation) | AdaBins · NYU-Depth-v2 test | δ1: 0.903 -> 0.888 | Hurt |
| Increase the number of bins above 256 (ablation) | AdaBins · NYU-Depth-v2 | Absolute Relative Error | No clear effect |
| Use attention after the decoder at higher spatial resolution rather than after the encoder at the bottleneck | AdaBins architecture · NYU-Depth-v2 | depth estimation performance | Helped |
Closest papers by meaning
- Neural RGB(r)D Sensing: Depth and Uncertainty From a Video Camera (CVPR 2019 · similarity 0.88)
- Differentiable Diffusion for Dense Depth Estimation From Multi-View Images (CVPR 2021 · similarity 0.87)
- Single Image Depth Estimation Trained via Depth From Defocus Cues (CVPR 2019 · similarity 0.87)
- DME: Unveiling the Bias for Better Generalized Monocular Depth Estimation (AAAI 2024 · similarity 0.85)
- PatchFusion: An End-to-End Tile-Based Framework for High-Resolution Monocular Metric Depth Estimation (CVPR 2024 · similarity 0.85)
- Improving Depth Completion via Depth Feature Upsampling (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.