Research map / Vision-language and multimodal / Human and embodied modeling
Location Sensitive Image Retrieval and Tagging
Raul Gomez, Jaume Gibert, Lluis Gomez, Dimosthenis Karatzas
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 |
|---|---|---|---|
| use multi-class softmax training instead of multi-label classification for hashtag supervision | ResNet-50 hashtag model · YFCC100M hashtag retrieval | P@10: 1.01 -> 14.07 | Improved |
| use multi-class softmax training instead of GloVe embedding regression | ResNet-50 hashtag model · YFCC100M hashtag retrieval | P@10: 7.02 -> 14.07 | Improved |
| add raw locations to LocSens (ablation) | LocSens · YFCC100M location sensitive retrieval, street (1 km) | P@10: 3.74 -> 4.83 | Improved |
| add raw locations to LocSens (ablation) | LocSens · YFCC100M location agnostic retrieval | P@10: 40.05 -> 32.74 | Hurt |
| add location dropout and progressive fusion instead of raw locations | LocSens · YFCC100M retrieval, continent (2500 km) | P@10: 28.42 -> 30.42 | Improved |
| add location dropout and progressive fusion instead of raw locations | LocSens · YFCC100M location agnostic retrieval | P@10: 32.74 -> 36.95 | Improved |
| add location dropout and progressive fusion instead of raw locations | LocSens · YFCC100M retrieval, street (1 km) | P@10: 4.83 -> 4.64 | Hurt |
| add location sampling with sigma=0.05 compared to sigma=1 (uninformative locations) (ablation) | LocSens · YFCC100M retrieval, continent (2500 km) | P@10: 28.40 -> 31.25 | Improved |
| lower sigma from 0.05 to 0 in location sampling (ablation) | LocSens · YFCC100M location agnostic retrieval | P@10: 39.80 -> 35.95 | Hurt |
| lower sigma from 0.05 to 0 in location sampling (ablation) | LocSens · YFCC100M retrieval, street (1 km) | P@10: 4.30 -> 4.83 | Improved |
| add location input (raw) to the tagging model (ablation) | LocSens vs zeroed locations · YFCC100M image tagging | A@1: 15.92 -> 28.10 | Improved |
| add location input (raw) to the tagging model (ablation) | LocSens vs zeroed locations · YFCC100M image tagging | A@10: 46.60 -> 68.21 | Improved |
Closest papers by meaning
- GeoRanker: Distance-Aware Ranking for Worldwide Image Geolocalization (NeurIPS 2025 · similarity 0.86)
- G3: An Effective and Adaptive Framework for Worldwide Geolocalization Using Large Multi-Modality Models (NeurIPS 2024 · similarity 0.86)
- Weatherproofing Retrieval for Localization with Generative AI and Geometric Consistency (ICLR 2024 · similarity 0.85)
- Scaling Image Geo-Localization to Continent Level (NeurIPS 2025 · similarity 0.85)
- Focus on Local: Finding Reliable Discriminative Regions for Visual Place Recognition (AAAI 2025 · similarity 0.85)
- Recognition through Reasoning: Reinforcing Image Geo-localization with Large Vision-Language Models (NeurIPS 2025 · 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.
