DeploySharp PaddleOCR and PP-Structure Models

Individual PP-OCR v4/v5/v6 core assets and converted PP-Structure assets. Download only the required model, dictionary, or generation-bundle files. Generated 2026-09-24. See paddleocr-release-catalog.json and SHA256SUMS for the complete manifest. Skipped conversion-blocked models are listed in the catalog.

Added 2026-09-24: PP-Chart2Table image-to-table generation bundle with four ONNX graphs and the official tokenizer assets. Download the required files listed under paddle-chart/pp-chart2table in the release catalog. Assets: pp-chart2table-vision-projector.onnx, pp-chart2table-token-embedding.onnx, pp-chart2table-text-prefill.onnx, pp-chart2table-text-decode-with-past.onnx, qwen.tiktoken, tokenizer_config.json, added_tokens.json.

Added 2026-10-26: experimental PaddleOCR-VL-1.6 five-graph ONNX Bundle (vision, embedding, decoder, head, derived prefill) plus tokenizer and manifest. Ten assets are individually downloadable; the updated paddleocr-release-catalog.json records their exact SHA-256 values and local names. This is verified for the pinned 1.6 Bundle only; backend and quality limits remain documented in DeploySharp.

Added 2026-10-27: separate OpenVINO-compatible SLANeXt wired/wireless ONNX graphs. The original graphs remain unchanged. Download slanext-wired-openvino-compat.onnx or slanext-wireless-openvino-compat.onnx individually; paddleocr-release-catalog.json and SHA256SUMS record their exact sizes and SHA-256 values. These derived graphs address the current OpenVINO Loop importer boundary, not table accuracy or TensorRT support.