Research map / Image generation / Autoregressive and latent generation
Local Features and Visual Words Emerge in Activations
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 |
|---|---|---|---|
| Add DSM spatial reranking | VGG16-GeM retrieval · ROxf medium | mAP: 44.8 -> 51.1 | Helped |
| Add DSM spatial reranking | VGG16-GeM retrieval · RPar medium | mAP: 65.7 -> 66.2 | Helped |
| Add spatially verified images as diffusion seeds | VGG16-GeM retrieval · ROxf medium | mAP: 48.4 -> 61.6 | Helped |
| Add DSM spatial reranking | ResNet101-GeM upsampled · RPar hard | mAP: 78.9 -> 79.3 | Hurt |
| Upsample activation maps (ablation) | ResNet101-GeM · ROxf hard | mAP: 38.5 -> 39.6 | Helped |
| Upsample activation maps (ablation) | ResNet101-MAC retrained · ROxf hard | mAP: 36.7 -> 35.6 | Hurt |
| Add supervised whitening (ablation) | VGG16-MAC retrained with DSM · ROxf medium | mAP: 58.2 -> 60.0 | Helped |
| Add DSM after supervised whitening (ablation) | ResNet101-MAC retrained, upsampled · ROxf medium | mAP: 63.9 -> 62.7 | Hurt |
| Add DSM before diffusion (ablation) | VGG16-MAC retrained · ROxf medium | mAP: 67.7 -> 72.0 | Helped |
| Add DSM before diffusion (ablation) | ResNet101-GeM upsampled · ROxf+R1M medium | mAP: 39.6 -> 41.9 | Helped |
| Use off-the-shelf network rather than retrieval-retrained network (ablation) | VGG16-MAC with DSM · ROxf medium | mAP: 51.1 -> 60.0 | Mixed |
Closest papers by meaning
- Kernelized Subspace Pooling for Deep Local Descriptors (CVPR 2018 · similarity 0.88)
- Guided Similarity Separation for Image Retrieval (NeurIPS 2019 · similarity 0.87)
- Local Descriptors Optimized for Average Precision (CVPR 2018 · similarity 0.87)
- Focus on Local: Finding Reliable Discriminative Regions for Visual Place Recognition (AAAI 2025 · similarity 0.86)
- SCENIR: Visual Semantic Clarity through Unsupervised Scene Graph Retrieval (ICML 2025 · similarity 0.85)
- DGC-GNN: Leveraging Geometry and Color Cues for Visual Descriptor-Free 2D-3D Matching (CVPR 2024 · similarity 0.84)
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.