TrOCR Base
Free.ai (self-hosted)
·
ocr
·
~300 Token kwa page
TrOCR Base bụ an OCR móòdù e mepụtara site na Microsoft. Ọrụ na Handwritten documents, low-resource scripts.. Self-hosted na Free.ai GPUs - na-agba ọsọ n'efu megide ụbọchị gị token pool (300 tokens Ihu). E wepụtara ya n'okpuru MIT — iji azụmahịa ekwenyela na Free.ai.
Jiri site na API
OpenAI-na-akpaghị aka REST API. Kewapụta kii nakwa kpọọ móòdù a n'ime sekọnd.
curl -X POST https://api.free.ai/v1/ocr/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"trocr-base","prompt":"your prompt here"}'Ajụjụ ndị a jụrụkarị
TrOCR Base na-ewepụ ngwe site na inyogo na PDFs. Bipụta ihuakwụkwọ skenètị̀, skịnètị̀, mọọbụ foto nke dọkumenti ahụ na TrOCR Base na-eziga ngwe ahụ — na-echekwa nkwụsị̀laịnụ na (n'ime engines nwere ike) táàbụ̀, fómùlà, nakwa nhazi.
TrOCR Base na-ejikwa ikiritị ndị dị iche iche - Latin, Cyrillic, CJK, Arabic, Devanagari, na ndị ọzọ. /ocr/ landing nwere ndesịta asụsụ zuru ezu maka engine.
Yabụ maka mbipụta-ụdị-n'aka-n'aka; cursive na shorthand dị njọ. TrOCR na nsụgharị-n'aka
Layout-aware engines (GOT-OCR, Marker, Florence-2) na-eziga Markdown/HTML table na LaTeX maka fomulụ. Plain-text engines na-eziga ngwe na-enweghị nhazi - hụ na mode nke mmepụta nke móòdù a.
TrOCR Base na-agba ọsọ na GPUs anyị site na pool gị n'ụbọchị; $5 → 200,000 tokens na-akwụ ụgwọ mgbe. ~300 tokens kwa peeji.
PNG, JPG, WebP, HEIC, BMP, nakwa otu na ọtụtụ ibe PDFs ruo 50 MB. A na-ewepụ ihe oyiyi ndị ahụ n'ụzọ ọbụla tupu a wepụ ha.
Nhazi nke akara bụ>98% na ihuakwụkwọ mbipụta dị ọcha, na-apụ na foto nke dị ala-reso, scans nke dị njọ, mọọbụ ihumkpụrụedemede na-adịghị asị. Tụnyere engines na /ocr/compare/ mgbe nhazi dị mkpa.
Ee — /batch/ na-anabata nsomebefaịlụ nke inyogo/PDFs. Nhazi ọbụla na-abanye na /account/?tab=history. Maka nchebe nsomebefaịlụ-tree jiri API.
Ya - POST faịlụ gị na /v1/ocr/ na model="TrOCR Base". Na-eziga JSON na ngwe + (n'oge a na-akwado ya) nhazi + okwu-nhazi bóòdù. /api/ nwere ndesịta zuru ezu.
Models na-echekwa onwe ha na-echekwa faịlụ gị na GPUs anyị; premium na-aga site na DPA. Dọkumenti a na-ehichapụ mgbe windo n'otu n'otu gasịrị. Anyị anaghị arụ ọrụ na init gị.
Ee - Free.ai na-enye ikike iji ngwe ahụ ebudatara. Ichọrọ ikike na dọkumenti isi.
1-5 sekọnd n'ihuakwụkwọ ọbụla. Layout-aware engines (taịbụl, mathematik) na-ewe ogologo oge. Multi-page PDFs na-egosipụta n'ụzọ linear.