PaddleOCR-Edge-v5

Latest PaddleOCR with edge-tuned recognition head. 80+ languages, single binary.

OCRApache-2.0INT8FP16ocrmultilingual

142K downloads 8.4K deploymentsUpdated Apr 4, 2028

Headline:18.4ms · Qualcomm QCS8550 · INT8

Overview Benchmarks6 Sim Results Deploy6 Files Discussion23

Cross-chip benchmark matrix

Every supported chip, in matched-pair runs from the Fo’c’sle HIL lab. Sortable by any column — click a header. Cells where the chip can’t run this model show Not supported.

Chip platform	Quant	Latency p50(ms)	Latency p95(ms)	Throughput	Acc. retention(%)	Power(W)	Memory(MB)	Tested by
Q Qualcomm QCS8550 48 TOPS · SoC	INT8	6.48	8.02	154 FPS	98.6	12.7	11	Fo’c’sle HIL
Apple Neural Engine (M4) 38 TOPS · SoC	FP16	10.2	12.2	98 FPS	99.4	7.2	22	Community
H Hailo-8 26 TOPS · M.2	INT8	12.0	14.2	83 FPS	97.3	2.4	11	Fo’c’sle HIL
π Raspberry Pi 5 + Hailo HAT 26 TOPS · HAT	INT8	12.2	15.6	82 FPS	97.8	6.8	11	Fo’c’sle HIL
Q Qualcomm QCS6490 12 TOPS · SoC	INT8	15.0	19.0	67 FPS	97.6	8.7	11	Fo’c’sle HIL
R Rockchip RK3588 6 TOPS · SoC	INT8	33.8	44.1	30 FPS	98.2	8.7	11	Community
H Hailo-10H 40 TOPS · M.2	Not supported
Q Snapdragon 8 Gen 3 NPU 45 TOPS · SoC	Not supported
N NVIDIA Jetson Orin Nano 40 TOPS · SoM	Not supported
N NVIDIA Jetson AGX Orin 275 TOPS · Module	Not supported
N NVIDIA Jetson Thor 2070 TOPS · Module	Not supported
A Ambarella CV5 16 TOPS · SoC	Not supported
A Ambarella CV72 32 TOPS · SoC	Not supported
M MediaTek Genio 700 4 TOPS · SoC	Not supported
G Google Coral Edge TPU 4 TOPS · USB	Not supported
I Intel Movidius Myriad X 4 TOPS · SoC	Not supported
A AMD Versal AI Edge VE2302 22 TOPS · SoC	Not supported

Leader per columnICDAR 2019 · 1280×720 doc scan · det+rec end-to-end

HIL conditions

All numbers measured on Fo’c’sle HIL rigs in Tel Aviv (primary), Munich (secondary), and Pittsburgh (robotics). Single-stream, batch-1, real preprocessing, real downstream consumer. p50/p95 are over 10,000-frame steady-state windows after a 30-second warm-up. Power draw is package power, not wall power. Memory footprint is the resident model + activations footprint at peak — not on-disk.

Submitted publisher numbers are accepted only if they reproduce within ±8% of an HIL-lab matched run on the same chip in the same input mode. Otherwise they live separately under the Discussion tab.