Research map / Detection and segmentation / Open-set and few-shot detection
On Equivariant and Invariant Learning of Object Landmark Representations
Zezhou Cheng, Jong-Chyi Su, Subhransu Maji
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 linear projection trained with spatial contrastive loss to hypercolumn (ablation) | Ours hypercolumn · MAFL different-identity matching (aligned) | mean pixel error: 6.16 -> 2.06 | Improved |
| add linear projection to hypercolumn (ablation) | Ours hypercolumn · MAFL different-identity matching (in-the-wild) | mean pixel error: 5.58 -> 3.03 | Improved |
| add linear projection to hypercolumn (ablation) | Ours hypercolumn · MAFL same-identity matching (aligned) | mean pixel error: 0.73 -> 0.71 | No clear effect |
| stack hypercolumn blocks #2-#5 instead of single block #4 (ablation) | Ours ResNet50 · AFLW_M landmark detection | inter-ocular distance error: 8.83 -> 6.99 | Improved |
| stack hypercolumn blocks #2-#5 instead of single block #4 (ablation) | Ours ResNet50 · MAFL landmark detection | inter-ocular distance error: 2.73 -> 2.44 | Improved |
| use deeper block 4 instead of block 1 features (ablation) | Ours ResNet50 · AFLW_M landmark detection | inter-ocular distance error: 24.20 -> 8.83 | Improved |
| add contrastive pretraining instead of random initialization (ablation) | ResNet50 hypercolumn · AFLW_M landmark detection | inter-ocular distance error: 16.74 -> 6.99 | Improved |
| add contrastive pretraining instead of ImageNet pretraining (ablation) | ResNet50 hypercolumn · AFLW_M landmark detection | inter-ocular distance error: 8.88 -> 6.99 | Improved |
| fine-tune whole network (ablation) | Ours block 4 ResNet50 · MAFL / AFLW_M landmark detection | inter-ocular distance error: 2.73 -> 2.81 (MAFL), 8.83 -> 7.80 (AFLW_M) | Mixed |
| replace NMF with learned projection for 64-D reduction (ablation) | Ours · AFLW_M landmark detection | inter-ocular distance error: 7.60 -> 7.21 | Improved |
| use ResNet50 instead of ResNet18 (ablation) | Ours hypercolumn · AFLW_M landmark detection | inter-ocular distance error: 8.59 -> 6.99 | Improved |
| train on 128x128 images instead of 96x96 (ablation) | Ours ResNet50 · CUB landmark detection | PCK: 68.63 -> 72.61 | Improved |
Closest papers by meaning
- Unsupervised Learning of Landmarks by Descriptor Vector Exchange (ICCV 2019 · similarity 0.90)
- Unsupervised Discovery of Object Landmarks as Structural Representations (CVPR 2018 · similarity 0.89)
- Unsupervised Learning of Object Landmarks through Conditional Image Generation (NeurIPS 2018 · similarity 0.88)
- Structuring Representation Geometry with Rotationally Equivariant Contrastive Learning (ICLR 2024 · similarity 0.88)
- Learning Generalizable Shape Completion with SIM(3) Equivariance (NeurIPS 2025 · similarity 0.87)
- Probing Equivariance and Symmetry Breaking in Convolutional Networks (NeurIPS 2025 · similarity 0.86)
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.
