Research map / Detection and segmentation / Open-set and few-shot detection
Unsupervised Learning of Landmarks by Descriptor Vector Exchange
James Thewlis, Samuel Albanie, Hakan Bilen, Andrea Vedaldi
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 DVE to 64D embedding (ablation) | equivariance baseline SmallNet · MAFL different-identity matching | pixel error: 5.68 -> 2.77 | Improved |
| add DVE to 16D embedding (ablation) | equivariance baseline SmallNet · MAFL different-identity matching | pixel error: 5.65 -> 2.79 | Improved |
| add DVE to 3D embedding (ablation) | equivariance baseline SmallNet · MAFL different-identity matching | pixel error: 2.89 -> 3.03 | Hurt |
| add DVE to 64D embedding (ablation) | equivariance baseline SmallNet · MAFL same-identity matching | pixel error: 1.25 -> 1.28 | No clear effect |
| increase embedding dimension from 3 to 64 without DVE (ablation) | equivariance baseline · MAFL different-identity matching | pixel error: 2.89 -> 5.68 | Hurt |
| increase embedding dimension from 3 to 64 without DVE (ablation) | equivariance baseline · MAFL same-identity matching | pixel error: 1.33 -> 1.25 | Improved |
| add DVE to 20D embedding (ablation) | equivariance baseline · Roboarm matching | matching error: 10.34 -> 1.25 | Improved |
| add DVE to 3D embedding (ablation) | equivariance baseline · Roboarm matching | matching error: 1.42 -> 1.41 | No clear effect |
| add transformations (flow-based warps) to DVE, 20D (ablation) | DVE · Roboarm matching | matching error: 1.42 -> 1.25 | Improved |
| add unsupervised finetuning on AFLW to Hourglass 64D (ablation) | DVE Hourglass · AFLW_M landmark regression | error: 8.15 -> 7.53 | Improved |
| add unsupervised finetuning on AFLW to SmallNet 3D (ablation) | DVE SmallNet · AFLW_M landmark regression | error: 11.82 -> 11.12 | Improved |
| add unsupervised finetuning to Hourglass 64D (ablation) | DVE Hourglass · 300W landmark regression | error: 4.65 -> 4.65 | No clear effect |
Closest papers by meaning
- On Equivariant and Invariant Learning of Object Landmark Representations (ICCV 2021 · similarity 0.90)
- Unsupervised Learning of Object Landmarks via Self-Training Correspondence (NeurIPS 2020 · similarity 0.88)
- Unsupervised Learning of Object Landmarks through Conditional Image Generation (NeurIPS 2018 · similarity 0.88)
- Unsupervised Learning of Object Landmarks by Factorized Spatial Embeddings (ICCV 2017 · similarity 0.88)
- Unsupervised Learning of Local Equivariant Descriptors for Point Clouds (TPAMI 2021 · similarity 0.87)
- Object landmark discovery through unsupervised adaptation (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.
