The AICrowd dataset has 90% training duplicates and 93% validation-to-training leakage; a perceptual hashing pipeline detects and mitigates these issues.
Deep learning for understanding satellite im- agery: An experimental survey
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2023 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Data Leakage Detection and De-duplication in Large Scale Geospatial Image Datasets
The AICrowd dataset has 90% training duplicates and 93% validation-to-training leakage; a perceptual hashing pipeline detects and mitigates these issues.