surya

工具介绍

OCR,布局分析,读取顺序,90+语言的表识别

核心特点

• •Accuracy - scores 83.3% on olmOCR-bench (top under 3B params)
• •Speed - throughput of 5 pages/s on an RTX 5090
• •Multilingual - scores 87.2% on an internal benchmark set of 91 languages (more here)
• •Layout analysis (table, image, header, etc.) with reading order
• •Table recognition (rows + columns)
• •NVIDIA GPU: Docker plus the NVIDIA Container Toolkit.
• •CPU / Apple Silicon: the llama-server binary from llama.cpp:
• •SuryaInferenceManager replaces FoundationPredictor. Same manager instance is shared across LayoutPredictor, RecognitionPredictor, TableRecPredictor.
• •Inspect the settings in surya/settings.py. You can override any setting via env var (e.g. SURYA_INFERENCE_BACKEND=vllm).
• •Text detection and OCR errors are separate models.