Inputs to provide
Original model, input and output definitions, calibration data, board, system image and target metrics.
AI Vision & Edge Intelligence
RK3588 / RK3576 model conversion and deployment
Adapt an existing model for an agreed Rockchip board, runtime and input pipeline.
Scoped deliverables
RKNN conversion, preprocessing and postprocessing, inference integration, build scripts and board test report.
Acceptance method
Compare model outputs, latency, memory use and stability before and after conversion on the named board.
Scope boundaries
Training, production firmware signing and third-party model rights require separate agreement.
From review to delivery
- 01Requirements and material review
- 02Proof of concept or prototype
- 03Development, integration and testing
- 04Deployment, pilot production and maintenance
We review materials before confirming scope, price, schedule and acceptance in writing. Metrics depend on agreed hardware, samples and operating conditions. Hardware, third-party licenses, certification, site work and ongoing operations are confirmed separately.
Discuss this service
We review the materials and scope before confirming a quotation and implementation schedule.