Research map / Detection and segmentation / Weakly supervised detection and segmentation
Self-produced Guidance for Weakly-supervised Object Localization
Xiaolin Zhang, Yunchao Wei, Guoliang Kang, Yi Yang, Thomas Huang
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 self-produced guidance (SPG-B and SPG-C) to SPG-plain | SPG-plain · ILSVRC val top-1 | top-1 localization error (%): 53.71 -> 51.40 | Improved |
| add self-produced guidance to SPG-plain | SPG-plain · ILSVRC val top-5 | top-5 localization error (%): 41.81 -> 40.00 | Improved |
| add self-produced guidance to SPG-plain | SPG-plain · CUB-200-2011 top-1 | top-1 localization error (%): 56.33 -> 53.36 | Improved |
| add self-produced guidance to SPG-plain | SPG-plain · ILSVRC val (GT-known) | GT-known localization error (%): 37.32 -> 35.31 | Improved |
| add two-stage cascade supervision of B1 by B2 (ablation) | SPG · ILSVRC val (GT-known) | GT-known localization error (%): 35.58 -> 35.31 | Improved |
| share second and third layers between B1 and B2 (ablation) | SPG · ILSVRC val (GT-known) | GT-known localization error (%): 36.31 -> 35.31 | Improved |
| add SPG-C auxiliary supervision (ablation) | SPG · ILSVRC val (GT-known) | GT-known localization error (%): 36.06 -> 35.31 | Improved |
| SPG-B alone without SPG-C versus plain network (ablation) | SPG without SPG-C · ILSVRC val (GT-known) | GT-known localization error (%): 37.32 -> 36.06 | Improved |
| repeat boxes from top predicted classes for top-5 (SPG*) | SPG · ILSVRC val top-5 | top-5 localization error (%): 40.00 -> 35.05 | Improved |
| repeat boxes from top predicted classes for top-5 (SPG*) | SPG · CUB-200-2011 top-5 | top-5 localization error (%): 42.28 -> 40.62 | Improved |
| combine with stronger classifier ResNet-50 instead of own classification | GoogLeNet-SPG · ILSVRC val top-1 | top-1 localization error (%): 51.40 -> 48.79 | Improved |
| combine with DPN-ensemble classifier instead of ResNet-50 | GoogLeNet-SPG · ILSVRC val top-1 | top-1 localization error (%): 48.79 -> 43.83 | Improved |
Closest papers by meaning
- Learning to Localize Objects with Noisy Labeled Instances (AAAI 2019 · similarity 0.88)
- Self-Supervised Object Localization with Joint Graph Partition (AAAI 2022 · similarity 0.88)
- Inter-Image Communication for Weakly Supervised Localization (ECCV 2020 · similarity 0.87)
- Deep Self-Taught Learning for Weakly Supervised Object Localization (CVPR 2017 · similarity 0.87)
- Spatial-Aware Token for Weakly Supervised Object Localization (ICCV 2023 · similarity 0.87)
- Attention-Based Dropout Layer for Weakly Supervised Object Localization (CVPR 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.
