Research map / 3D vision / Human pose estimation
ScaleNet: A Shallow Architecture for Scale Estimation
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 |
|---|---|---|---|
| Apply ScaleNet before feature extraction | SIFT on the Megadepth custom split · Megadepth custom split | Pose AUC at 5°: 9.8 -> 12.9; +32% | Helped |
| Apply ScaleNet before feature extraction | SP + SuperGlue on the Megadepth custom split · Megadepth custom split | Pose AUC at 5°: 16.5 -> 22.5; +36% | Helped |
| Rescale one image before dense correspondence estimation | DGC-Net on the Megadepth Hard split · Megadepth, Hard split | PCK-5: 4.5 -> 20.1; +347% | Helped |
| Rescale one image before dense correspondence estimation | GLU-Net on the Megadepth Hard split · Megadepth, Hard split | PCK-5: 10.8 -> 26.8; +148% | Helped |
| Apply ScaleNet image rectification | SuperPoint reconstruction on IMC test scenes · IMC test scenes | Track length: 10.47 -> 12.71 | Helped |
| Apply ScaleNet image rectification | SuperPoint reconstruction on IMC test scenes · IMC test scenes | Reprojection error: 1.17 -> 1.26 | Hurt |
| Use inverse-scale consistency check (ablation) | ScaleNet on the Megadepth custom validation split · Megadepth custom validation split | Pose AUC at 5°: 8.4 -> 8.7 | Helped |
| Add ASPP to VGG with self-correlation (ablation) | ScaleNet on the Megadepth custom validation split · Megadepth custom validation split | Pose AUC at 5°: 6.2 -> 8.4 | Helped |
| Add self-correlation to VGG with ASPP (ablation) | ScaleNet on the Megadepth custom validation split · Megadepth custom validation split | Pose AUC at 5°: 7.3 -> 8.4 | Helped |
| Compute soft scale in logarithmic rather than natural space (ablation) | ScaleNet on the Megadepth custom validation split · Megadepth custom validation split | Pose AUC at 5°: 7.2 -> 8.4 | Helped |
| Predict a scale distribution rather than regress one scale value (ablation) | ScaleNet on the Megadepth custom validation split · Megadepth custom validation split | Pose AUC at 5°: 4.9 -> 8.4 | Helped |
| Use soft assignment instead of hard assignment (ablation) | ScaleNet on the Megadepth custom validation split · Megadepth custom validation split | Pose AUC at 5°: 7.9 -> 8.4 | Helped |
Closest papers by meaning
- ScaleNet - Improve CNNs through Recursively Rescaling Objects (AAAI 2020 · similarity 0.90)
- Data-Driven Neuron Allocation for Scale Aggregation Networks (CVPR 2019 · similarity 0.89)
- Revisiting ResNets: Improved Training and Scaling Strategies (NeurIPS 2021 · similarity 0.88)
- Scene Recognition With CNNs: Objects, Scales and Dataset Bias (CVPR 2016 · similarity 0.88)
- An Analysis of Scale Invariance in Object Detection SNIP (CVPR 2018 · similarity 0.87)
- Deep Scale-spaces: Equivariance Over Scale (NeurIPS 2019 · similarity 0.87)
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.