Research map / Image generation / Autoregressive and latent generation
Data Leakage Detection and De-duplication in Large Scale Geospatial Image Datasets
Yeshwanth Kumar Adimoolam, Charalambos Poullis, Melinos Averkiou
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 |
|---|---|---|---|
| account for rotations and flips by hashing augmented copies (ablation) | perceptual hashing leakage detection · AICrowd val vs train | leakage rate: 93.45% -> 97.04% | Improved |
| use average hashing instead of perceptual hashing (ablation) | leakage detection · AICrowd official train vs val | duplicates detected: 33.9% -> 34.9% | No clear effect |
| use average hashing instead of perceptual hashing (ablation) | leakage detection · AICrowd official val vs official test | duplicates detected: 17.7% -> 18.6% (val); 17.6% -> 18.7% (test) | No clear effect |
| apply hashing pipeline to INRIA dataset | perceptual hashing · INRIA Aerial Image Labelling | overlap (%): max 25 x 10-3% overlap | No clear effect |
| apply hashing pipeline to SpaceNet 2 | perceptual hashing · SpaceNet 2 Building Detection v2 | overlap (%): about 0.85% to 1.54% overlap | No clear effect |
| de-duplicate and remove leakage from AICrowd | perceptual hashing · AICrowd train split | images: 280,741 -> 29,338 -> 15,392 | Improved |
| train on leaked AICrowd splits | PolyWorld, HiSup · AICrowd val | qualitative overfitting | Hurt |
Closest papers by meaning
- Localization Is All You Evaluate: Data Leakage in Online Mapping Datasets and How to Fix It (CVPR 2024 · similarity 0.83)
- Benchmarking Object Detectors under Real-World Distribution Shifts in Satellite Imagery (CVPR 2025 · similarity 0.81)
- Learnable Earth Parser: Discovering 3D Prototypes in Aerial Scans (CVPR 2024 · similarity 0.81)
- SELECT: A Large-Scale Benchmark of Data Curation Strategies for Image Classification (NeurIPS 2024 · similarity 0.81)
- Large Scale Dataset Distillation with Domain Shift (ICML 2024 · similarity 0.80)
- FADRM: Fast and Accurate Data Residual Matching for Dataset Distillation (NeurIPS 2025 · similarity 0.80)
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.
