ruisv/bcdl-ppocr-v6

PP-OCRv6 for RDK S100/S100P — detection + recognition + angle

Compiled BPU models (.hbm) for the D-Robotics RDK S100 / S100P, ready to load — no ONNX export, no calibration, no hb_compile. Built and measured with BCDL, a C++17 inference and media library for the RDK S-series with Python bindings.

Upstream: PaddleOCR PP-OCRv6 medium

Redistributable, including commercially. The licence chain was checked on the code, the pretrained weights it started from, and the data it was trained on — all three, because a permissive repository badge does not by itself say anything about the weights. See Licence.

Files

file	what it is
ppocrv6_medium_det_960x960.hbm	text detection, 960x960 — 22.4 MB
ppocrv6_medium_rec_int16_48x320.hbm	recognition, 48x320, all-int16 — 24.0 MB
ppocrv5_lcnet_cls_80x160.hbm	textline angle 0/180 (from v5 — v6 ships none) — 2.3 MB
ppocr_keys_v6_18710.txt	dictionary — required by the recogniser

Measured on an S100P

stage	latency	throughput
detection	8.55 ms	117 FPS
recognition	2.22 ms	451 FPS
angle classifier	0.37 ms	2710 FPS

hrt_model_exec perf, one thread, minimum of three runs, on a board first gated against its own previously recorded numbers. BPU time only — CPU pre/post-processing is on top and is listed per task in BCDL's benchmark results.

Use it

conda install -c https://mirrors.ruis.ai/conda -c conda-forge bcdl

import bcdl
engine = bcdl.Engine("ppocrv6_medium_det_960x960.hbm")
print(engine.input_shape(0), engine.output_shape(0))

Each task has a decoder in BCDL that turns those raw outputs into boxes, keypoints, masks, disparity or text — see the Python API (中文).

What to know before deploying

The recogniser is all-int16 on purpose. On the S100P it roughly halves the character error of the compiler's default mixed-precision int8 build and is about 2x faster, because a uniform int16 graph carries no int8-int16 requant on its internal boundaries.

The dictionary is not optional and it is version-specific. v6 recognises 18710 classes against v5's 18385; pairing a model with the wrong key file decodes to plausible-looking wrong characters rather than failing. The right one ships here.

For text lines longer than about 6.7:1 there is also a 960-wide recogniser in the conversion recipe; this package carries the 320 build.

Licence

Apache-2.0, on the code and on every model card.

BCDL itself is Apache-2.0 and is unrelated to these terms — it is a general-purpose runtime that loads any .hbm. The licence above constrains these weights and this compiled artefact.

The conversion recipe — ONNX export, calibration, hb_compile config and the acceptance numbers — is public in bcdl-model-zoo, so this build can be reproduced or retargeted rather than taken on trust.