PaddleOCR (Baidu) — AI 뉴스 및 모델

모델 🇺🇸

PP-OCRv6 on Hugging Face: Text Recognition in 50 Languages with Models from 1.5M to 34.5M Parameters

The PaddleOCR team has released PP-OCRv6, a new family of universal OCR models supporting 50 languages. The models are available in three sizes (tiny, small, medium) with parameter counts ranging from 1.5M to 34.5M. Detection and recognition are significantly improved compared to the previous PP-OCRv5_server version.

PaddleOCR (Baidu)

Hugging Face blog27.07 · 11:04

애플리케이션 🇷🇺

How Lenta Tech Searched for Price Tags on Video from a Moving Robot: Results of a Video Analytics Hackathon

The Lenta Tech team (the IT brand of the Lenta Group) shared their experience of holding a hackathon on recognizing price tags from video recorded by a moving robot. The article describes the solution pipeline: from price tag detection using YOLO to data extraction via OCR, QR codes, and a catalog. Key takeaways: quality is achieved not by a single model, but by proper handling of the frame sequence (tracking, selecting the best frame) and combining OCR, QR, and the catalog.

Ultralytics

PaddleOCR (Baidu)

OpenCV

EasyOCR (Jaided AI)

Tesseract OCR (Google)

Habr — хаб ИИ24.07 · 02:02