All pull requests

Pull requests

• [text recognition] fix: group accented Latin letters into a single word box

  #5222·li-lizhe opened Oct 13, 2026·

  ·
  1

• Add opt-in width_scale / width_limit to text recognition resize

  #5215·Ahmed-Sinkeat opened Oct 6, 2026·

  ·
  1

• security: restrict text detection scale expressions

  #5212·zhengguangzhuo opened Sep 29, 2026·

  ·
  1

• fix(deps): detect any distribution that provides cv2

  #5203·aeborisov opened Sep 8, 2026·

  ·
  1

• Fix save_to_markdown overwriting images with same coordinates on different pages

  #5195·jaideeppyne opened Aug 18, 2026·

  ·
  1

• Remove xformers dependency from genai server plugin (no runtime usage; unblocks aarch64 CUDA deploy)

  #5194·BreezeHavana opened Aug 17, 2026·

  ·

• Fix duplicate Markdown images across pages

  #5190·DresdenGman opened Aug 13, 2026·

  ·
  1

  2

• [Performance] Avoid redundant CTC logits copy during text recognition

  #5182·1443858742 opened Jul 28, 2026·

  ·
  1

• fix(ocr): make UVDoc output C-contiguous

  contributor
  #5179·1443858742 opened Jul 26, 2026·

  ·
  2

• fix: apply Heading style from BASE_STYLE_MAP level in Word output

  contributor
  #5174·Shengshenlan opened Jul 16, 2026·

  ·
  1

  2

• fix(paddleocr_vl): make repetition truncation catch multi-line and cap-truncated output

  contributor
  #5170·subin9 opened Jul 13, 2026·

  ·
  2

• Add headless OpenCV extras (*-headless)

  contributor
  #5169·isaiah-harville opened Jul 1, 2026·

  ·
  2

• fix(ocr): support Japanese kana word boxes

  contributor
  #5168·Ikwus opened Jun 28, 2026·

  ·
  1

  2

• Register missing korean/el/th PP-OCRv5_mobile_rec for HPI (gpu_cuda12/paddle311)

  contributor
  #5163·ddoron9 opened Jun 24, 2026·

  ·
  2

• fix(layout_parsing): validate chart and formula model settings

  contributor
  #5160·lixiyu opened Jun 14, 2026·

  ·
  3

• Fix max_box_w calculation bug in extract_polygon_points_by_masks

  contributor
  #5147·WenbinHou opened May 31, 2026·

  ·
  3

• feat: warn on large text detector inputs

  contributor
  #5146·hoobnn opened May 29, 2026Contributor·

  ·
  1

• docs: fix invalid OCR arguments and PP-ChatOCR config paths

  contributor
  #5145·hoobnn opened May 29, 2026Contributor·

  ·
  1

• (Fixes PaddlePaddle/PaddleOCR#18064) fix: guard against KMeans n_clusters=0 crash in table recognition pipeline

  contributor
  #5140·Mane2305 opened May 27, 2026·

  ·
  1

  3

• test: CI trigger verification

  contributor
  #5133·kjagsdq opened May 24, 2026·

  ·
  1

• [ROCm/DCU] Remove BF16 workarounds: is_bfloat16_available, delete_pass, _keep_in_fp32_modules

  contributorPaddlePaddle Hackathon
  #5112·oldzhu opened Apr 23, 2026·

  ·
  2

  

• fix(table): escape html-sensitive OCR text in table markdown output

  contributor
  #5101·jimmyzhuu opened Apr 20, 2026·

  ·
  3

  

• [ROCm] Remove BF16 workarounds for PaddleOCR-VL on HIP

  contributorPaddlePaddle Hackathon
  #5096·austin1997 opened Apr 18, 2026·

  ·
  2

• [ROCm] Remove BF16 workaround now that Paddle framework supports HIP BF16

  contributorPaddlePaddle Hackathon
  #5082·Lau-JW opened Apr 13, 2026·3 tasks done·

  ·
  2

  

• fix(doc_vlm): remove ROCm BF16 _keep_in_fp32_modules workaround in PaddleOCR-VL

  contributorPaddlePaddle Hackathon
  #5077·fchange opened Apr 4, 2026·

  ·
  2