PaddleOCR-VL
Free.ai (self-hosted)
·
ocr
·
~300 tokens per page
PaddleOCR-VL ke a Mofuta oa OCR e hahiloeng ke PaddlePaddle. E ka ba e le Fast multilingual image / scan to Markdown; tables, formulas, charts; 109 languages.. E na le li-Free.ai GPUs tse nang le li-server tse 100 tse nang le li-server tse 100 tse nang le li-server tse 100. E lokollotsoe ka Apache 2.0 — ho sebelisoa ka khoebo ho lumelloa ka Free.ai.
Ho sebelisa ka API
REST API e lumellanang le OpenAI. E etsa konopo'me e bitsa mofuta ona ka metsotsoana.
curl -X POST https://api.free.ai/v1/ocr/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"paddleocr-vl","prompt":"your prompt here"}'Lipotso tse botsoang khafetsa
PaddleOCR-VL e tlosa mongolo ho li-images le li-PDFs. Kopitsa leqephe le scanned, skrini, kapa setšoantšo sa lengolo-tsoibila'me PaddleOCR-VL e khutlisa mongolo — e boloka li-line breaks le (ka li-engine tse ka khonehang) li-table, li-formula, le sebopeho.
PaddleOCR-VL e sebetsana le li-scripts tse ngata — Latin, Cyrillic, CJK, Arabic, Devanagari, le tse ling. /ocr/ landing e na le lethathamo le felletseng la lipuo ka enjene.
E-na le ho ngola ka letsoho ka mokhoa o hlophisitsoeng; cursive le shorthand li thata. TrOCR le li-engine tse nang le li-transformer tse ncha li ntle haholo ho feta Tesseract ea khale ho ngola ka letsoho.
Li-engine tse tsebang ho hlophisa (GOT-OCR, Marker, Florence-2) li khutlisa li-table tsa Markdown/HTML le LaTeX bakeng sa li-formula. Li-engine tsa mongolo o hlakileng li khutlisa mongolo o se nang sebopeho — hlahloba mokhoa oa ho tsoa oa mofuta ona.
PaddleOCR-VL e sebetsa ka GPUs ea rona ho tloha ho pholletsa le letsatsi la hau la mahala; $5 → 200,000 tokens tse tšehelitsoeng ka mor'a. ~300 tokens ka leqephe.
PNG, JPG, WebP, HEIC, BMP, le li-PDF tse nang le li-page tse ngata ho fihlela ho 50 MB. Li-page tse otlolohileng kapa tse otlolohileng li ntlafatsoa ka ho toba pele ho ho tsoa.
Ho nepahala ha li-character ho ka ba>98% ka li-page tse ngotsoeng ka nepo, ho fella ka lifoto tse nang le bophahamo bo tlase, li-scan tse sa nepahale haholo, kapa li-fonti tse sa tloaelehang. Bapisa li-engine ho /ocr/compare/ ha ho tluoa tabeng ea ho nepahala.
Ee — /batch/ e amohela foldara ea litšoantšo/PDFs. Ho tlosoa ka bobeli ho tla ka /account/?tab=history. Ho boloka moqhaka oa foldara sebelisa API.
E-na — POST faele ea hau ho /v1/ocr/ le model="PaddleOCR-VL". E khutlisa JSON le mongolo + (ha e tšehetsoa) moralo + li-boxes tsa ho qeta ka lentsoe. /api/ e na le ho ngolisoa ka ho feletseng.
Li-model tse nang le ho amohela ka bomong li boloka faele ea hau ho GPUs tsa rona; premium e tsamaea ka DPA. Litokomane li tlosoa ka mor'a ho arolelana-fensetereng. Re sa koetlise ka li-input tsa hau.
Ee — Free.ai e fana ka ts'ebeliso ea khoebo ea mongolo o kopantsoeng. U hloka litokelo tsa ho kenya ho lifaele tsa mohloli.
1–5 lisegunete ka leqephe ka kakaretso. Li-engine tse nang le li-layout (tekanyo, lipalo) li nka nako e telele. Li-PDF tse nang le li-page tse ngata li theha ka mokhoa o hlakileng.