MediaPipe vs Yandex Vision OCR

MediaPipe gives developers libraries, tools, ready-to-run models and customizable APIs for adding AI and machine learning to applications. Its vision tasks include face and hand landmarks, gesture recognition, image classification and segmentation, object detection, and pose landmarks. Text tasks cover language detection, classification, embeddings, proofreading and summarization, with audio classification also available. MediaPipe Tasks supports Android, Web/JavaScript, Python and iOS; its cross-platform APIs are described as requiring as few as five lines of code for inference. The optimized on-device pipelines use CPU, GPU and TPU acceleration for real-time use cases. Developers can try web samples in MediaPipe Studio with text, webcam input or uploaded images. Model Maker supports transfer learning for compatible models, but it is deprecated and cannot change a model's original task. Solution API inputs are processed on-device and not sent to Google; usage and performance metrics are sent to Google. The framework can be built from source for Linux, macOS, Windows and Docker, with some configurations marked experimental.

Yandex Vision OCR recognizes printed and handwritten text in images and PDF files. Its models cover general page text, multi-column layouts, handwriting, tables, Markdown text, and mathematical formulas. Document-template models can recognize passports, driver licenses, vehicle registration certificates, and license plates, and return template fields. The service is accessed through gRPC or REST APIs. Results arrive as JSON Lines and can include recognized text and language, page, block, line, word, and image positions, plus layout types such as titles, headers, footnotes, and lists. Supported formats are JPEG, PNG, and PDF. The API accepts files up to 10 MB and images up to 20 megapixels; asynchronous PDF requests can cover up to 200 pages. Results remain on the server for three days. Each successful image or PDF-page recognition counts as one billed unit, while server errors and incorrectly configured requests are not charged. Listed rates range from free for printed text and license plate recognition to $0.01, paid once for several document, handwriting, and table models, billed per unit without VAT. There is no free plan.