Research map / Detection and segmentation / CNN features
Presence-Only Geographical Priors for Fine-Grained Image Classification
Oisin Mac Aodha, Elijah Cole, Pietro Perona
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 |
|---|---|---|---|
| combine spatio-temporal prior with image classifier | InceptionV3 · iNat2017 val | top-1 accuracy: 63.27 -> 69.60 | Improved |
| combine spatio-temporal prior with image classifier | InceptionV3 · iNat2018 val | top-1 accuracy: 60.20 -> 72.68 | Improved |
| combine prior with classifier on larger 520x520 images | InceptionV3 · iNat2018 val | top-1 accuracy: 66.18 -> 77.49 | Improved |
| combine prior with classifier | InceptionV3 · YFCC test | top-1 accuracy: 50.15 -> 50.70 | No clear effect |
| add date input to the prior (ablation) | Ours · iNat2017 val | top-1 accuracy: 69.34 -> 69.60 | Improved |
| add date input to the prior (ablation) | Ours · iNat2018 val | top-1 accuracy: 72.41 -> 72.68 | Improved |
| add date input to the prior (ablation) | Ours · NABirds test (simulated) | top-1 accuracy: 81.15 -> 81.50 | Improved |
| add sin/cos wrap encoding of inputs (ablation) | Ours · iNat2017 val | top-1 accuracy: 69.48 -> 69.60 | Improved |
| add sin/cos wrap encoding of inputs (ablation) | Ours · iNat2018 val | top-1 accuracy: 72.12 -> 72.68 | Improved |
| add photographer embedding (ablation) | Ours · iNat2017 val | top-1 accuracy: 69.39 -> 69.60 | Improved |
| add photographer embedding (ablation) | Ours · iNat2018 val (299x299) | top-1 accuracy: 72.84 -> 72.68 | Hurt |
| add photographer embedding (ablation) | Ours · iNat2018 val (520x520) | top-1 accuracy: 77.64 -> 77.49 | Hurt |
Closest papers by meaning
- A Large-Scale Study on Unsupervised Spatiotemporal Representation Learning (CVPR 2021 · similarity 0.85)
- In Defense of Image Pre-training for Spatiotemporal Recognition (ECCV 2022 · similarity 0.85)
- Geography-Aware Self-Supervised Learning (ICCV 2021 · similarity 0.85)
- CSP: Self-Supervised Contrastive Spatial Pre-Training for Geospatial-Visual Representations (ICML 2023 · similarity 0.85)
- Divide&Classify: Fine-Grained Classification for City-Wide Visual Geo-Localization (ICCV 2023 · similarity 0.85)
- Learning to Align Generative Appearance Priors for Fine-grained Image Retrieval (NeurIPS 2026 · 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.
