Research map / Detection and segmentation / Weakly supervised detection and segmentation
WILDCAT: Weakly Supervised Learning of Deep ConvNets for Image Classification, Pointwise Localization and Segmentation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use multiple modalities instead of M=1 (ablation) | WILDCAT · VOC 2007, VOC 2012 Action, and MIT67 | classification score: 89.0 -> 92.5; 78.9 -> 83.2; 69.6 -> 73.1 | Helped |
| increase modality count beyond the best-performing setting (ablation) | WILDCAT · VOC 2007, VOC 2012 Action, and MIT67 | classification score: 92.5 -> 92.0; 83.2 -> 82.7; 73.1 -> 72.9 | Hurt |
| use architecture conv5 + conv + pooling rather than conv5 + pooling + conv (ablation) | WILDCAT · VOC 2007, VOC 2012 Action, and MIT67 | classification score: 87.3 -> 89.0; 77.5 -> 78.9; 68.1 -> 69.6 | Helped |
| use α=0.7 instead of α=1 for weighting negative evidence (ablation) | WILDCAT · VOC 2007, VOC 2012 Action, and MIT67 | classification score: 89.0 -> 90.3; 78.9 -> 80.9; 69.6 -> 71.3 | Helped |
| use M=4 instead of M=1 (ablation) | WILDCAT · VOC 2007, VOC 2012 Action, and MIT67 | classification score: 89.0 -> 91.6; 78.9 -> 82.1; 69.6 -> 72.0 | Helped |
| combine α=0.7 and M=4 (ablation) | WILDCAT · VOC 2007, VOC 2012 Action, and MIT67 | classification score: 91.6 -> 92.0; 82.1 -> 82.9; 72.0 -> 72.8 | Helped |
| add FC-CRF post-processing (ablation) | WILDCAT · VOC 2012 | mean IoU: 39.2 -> 43.7 | Helped |
| include minimum-scoring regions alongside maximum-scoring regions (ablation) | WILDCAT · VOC 2012 classification and localization | score: +1.6 pt | Helped |
Closest papers by meaning
- Revisiting Dilated Convolution: A Simple Approach for Weakly- and Semi-Supervised Semantic Segmentation (CVPR 2018 · similarity 0.87)
- NetVLAD: CNN Architecture for Weakly Supervised Place Recognition (CVPR 2016 · similarity 0.87)
- CAM Back Again: Large Kernel CNNs from a Weakly Supervised Object Localization Perspective (CVPR 2024 · similarity 0.87)
- WildCat: Near-Linear Attention in Theory and Practice (ICML 2026 · similarity 0.86)
- Weakly Supervised Deep Detection Networks (CVPR 2016 · similarity 0.86)
- Universal Weakly Supervised Segmentation by Pixel-to-Segment Contrastive Learning (ICLR 2021 · 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.