1000 synthetic printed-text line images with ground-truth transcriptions, sampled from a larger locally-held Tamil OCR training corpus. This is a benchmark/sample release, not the full training set.

Data fields

Field	Description
`file_name`	relative path to the image (`images/...`)
`text`	ground-truth transcription
`category`	`tamil_only` / `english_only` / `mixed` / `numeric_and_symbols`
`length_bucket`	`short` / `medium` / `long`, by character count

Category distribution (this sample)

Category	Count
`tamil_only`	963
`english_only`	30
`numeric_and_symbols`	7

Length-bucket distribution (this sample)

Bucket	Count
`medium`	455
`long`	284
`short`	261

Note: this language currently has no mixed (native+Latin in the same line) samples.

A note on data quality and how to help

This corpus was reviewed by someone who cannot personally read or verify Tamil.

If you notice anything wrong — a mistranscription, an odd or non-natural sentence, wrong script/font rendering, or anything else that looks off — please open a thread in this repo's Community tab (Discussions) and mention the file_name and what's wrong. Real corrections from people who can actually read Tamil are exactly what's needed to improve this further, and are genuinely welcome.

License

MIT — see LICENSE.

Downloads last month: 342

Total file size:

16.9 MB