Research map / Detection and segmentation / Weakly supervised detection and segmentation
Hide-And-Seek: Forcing a Network to Be Meticulous for Weakly-Supervised Object and Action Localization
Krishna Kumar Singh, Yong Jae Lee
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 |
|---|---|---|---|
| hide random 32x32 patches during training | AlexNet-GAP · ILSVRC val | GT-known Loc: 54.90 -> 58.75 | Improved |
| hide random 32x32 patches during training | GoogLeNet-GAP · ILSVRC val | Top-1 Loc: 43.60 -> 45.21 | Improved |
| hide patches during training | AlexNet-GAP · ILSVRC val | Top-1 Clas: 60.23 -> 57.94 | Hurt |
| use mixed patch sizes (16,32,44,56 or none) per image (ablation) | AlexNet-HaS · ILSVRC val | Top-1 Loc: 37.34 -> 37.65 | Improved |
| ensemble HaS models of several patch sizes instead of five GAP models (ablation) | AlexNet · ILSVRC val | GT-known Loc: 56.91 -> 60.14 | Improved |
| hide patches instead of image-level 50% dropout (ablation) | AlexNet · ILSVRC val | Top-1 Loc: 31.68 -> 37.65 | Improved |
| use global max pooling with Hide-and-Seek instead of GAP baseline (ablation) | AlexNet · ILSVRC val | GT-known Loc: 50.40 -> 59.27 | Improved |
| max pooling instead of average pooling under Hide-and-Seek (ablation) | AlexNet-HaS-56 · ILSVRC val | GT-known Loc: 58.43 -> 59.27 | Improved |
| hide patches in conv1 feature map (size 11) (ablation) | AlexNet-GAP · ILSVRC val | GT-known Loc: 54.90 -> 58.33 | Improved |
| raise hiding probability from 50% to 75% (ablation) | AlexNet-HaS-56 · ILSVRC val | Top-1 Loc: 37.34 -> 34.21 | Hurt |
| lower hiding probability from 50% to 25% (ablation) | AlexNet-HaS-56 · ILSVRC val | GT-known Loc: 58.43 -> 57.49 | Hurt |
| hide frame segments during training | C3D action classifier · THUMOS 2014 val (IoU 0.5) | mAP: 6.11 -> 6.84 | Improved |
Closest papers by meaning
- Weakly-Supervised Action Localization With Background Modeling (ICCV 2019 · similarity 0.89)
- Bridging the Gap Between Classification and Localization for Weakly Supervised Object Localization (CVPR 2022 · similarity 0.88)
- Weakly Supervised Object Localization and Detection: A Survey (TPAMI 2021 · similarity 0.88)
- Evaluating Weakly Supervised Object Localization Methods Right (CVPR 2020 · similarity 0.88)
- Pointly-Supervised Action Localization (IJCV 2018 · similarity 0.88)
- RELOCATE: A Simple Training-Free Baseline for Visual Query Localization Using Region-Based Representations (CVPR 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.
