FlyOCR creator describes a PDF reader modeled on a fruit fly brain

FlyOCR’s creator says the project uses the full MaleCNS v1.0 fruit fly connectome—a map of neural connections—to read PDF images. It separates individual characters, maps their pixels to receptor activations, runs simplified neural dynamics through the circuit and decodes the resulting signals into text. The creator reports 87% accuracy on 1,700-plus sampled characters and digits. In a test on a Microsoft 10-K filing, they report about 86% on the balance-sheet heading and say the system was largely able to read the numeric values correctly.