Research map / Detection and segmentation / Weakly supervised detection and segmentation
Unsupervised Object Localization with Representer Point Selection
Yeonghwan Song, Seokwoo Jang, Dina Katabi, Jeany Son
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 |
|---|---|---|---|
| replace class-agnostic activation map with representer point selection inference (ablation) | MoCo v2 ResNet50 · ImageNet-1K GT-Known Loc | GT-Known Loc: 61.19 -> 69.10 | Improved |
| replace class-agnostic activation map with representer point selection inference (ablation) | MoCo v2 ResNet50 · CUB-200-2011 GT-Known Loc | GT-Known Loc: 86.93 -> 96.67 | Improved |
| apply representer point inference on top of trained C2AM (ablation) | C2AM · CUB-200-2011 GT-Known Loc | GT-Known Loc: 89.90 -> 96.86 | Improved |
| apply representer point inference on top of DAOL (ablation) | DAOL · ImageNet-1K GT-Known Loc | GT-Known Loc: 70.27 -> 70.97 | Improved |
| use a self-supervised backbone instead of supervised pretraining (ablation) | Representer point selection · ImageNet-1K GT-Known Loc | GT-Known Loc: 63.22 -> 66.89 | Improved |
| compute foreground predictor from 0.1% instead of 100% of the data (ablation) | Representer point selection (ViT-S) · ImageNet-1K GT-Known Loc | GT-Known Loc: 73.65 -> 73.67 | No clear effect |
| compute foreground predictor on CIFAR10 instead of the target dataset (ablation) | Representer point selection (ViT-S) · CUB-200-2011 GT-Known Loc | GT-Known Loc: 91.62 -> 92.73 | No clear effect |
| transfer foreground predictor across datasets without retraining | Representer point selection vs C2AM · CUB-200-2011 GT-Known Loc, trained on CIFAR10 | GT-Known Loc: 73.18 -> 94.24 | Improved |
| use ViT-S DINO backbone instead of ResNet50 MoCo v2 (ablation) | Representer point selection · CUB-200-2011 GT-Known Loc | GT-Known Loc: 96.67 -> 91.61 | Hurt |
| use a class-specific predictor w_c for ImageNet instead of class-agnostic (ablation) | Representer point selection · ImageNet-1K GT-Known Loc | GT-Known Loc: 73.65 -> 74.44 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- TokenCut — TokenCut: Segmenting Objects in Images and Videos With Self-Supervised Transformer and Normalized Cut (TPAMI 2023) · better on 2 of 2 shared results
- C2AM — C2AM: Contrastive Learning of Class-Agnostic Activation Map for Weakly Supervised Object Localization and Semantic Segmentation (CVPR 2022) · better on 2 of 2 shared results
- LCTR — LCTR: On Awakening the Local Continuity of Transformer for Weakly Supervised Object Localization (AAAI 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Unsupervised Object Localization in the Era of Self-Supervised ViTs: A Survey (IJCV 2024 · similarity 0.91)
- Unsupervised Open-Vocabulary Object Localization in Videos (ICCV 2023 · similarity 0.89)
- PsyNet: Self-Supervised Approach to Object Localization Using Point Symmetric Transformation (AAAI 2020 · similarity 0.89)
- Unsupervised Object Localization: Observing the Background To Discover Objects (CVPR 2023 · similarity 0.89)
- Self-Supervised Object Localization with Joint Graph Partition (AAAI 2022 · similarity 0.89)
- Weakly Supervised Object Localization and Detection: A Survey (TPAMI 2021 · similarity 0.89)
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.
