3 ms·
You can roughly classify the images first (we already have very good AI model for this) and use humans as an additional layer. If two images don't get the same
by raincole 4y ago
You can roughly classify the images first (we already have very good AI model for this) and use humans as an additional layer. If two images don't get the same top-3 labels from the classifier, the chance that they're duplicates is neglectable.