Before an SDK evaluation
Provide the board model, Linux distribution, board support package (BSP), RKNN Runtime version, available memory and image format. State the expected video stream count, frames processed per second per stream, and the acceptance goals for the application.
- Target board and runtime versions
- Image resolution, frame processing rate and concurrency
- Evaluation goals and acceptance criteria
Face feature extraction
Image or decoded video frame → face detection → face alignment → feature extraction → feature vector. Alignment uses the landmarks supported by the model. The package specifies preprocessing, output format and model version.
Read the Face SDK overviewBody feature extraction
Image or decoded video frame → person detection → person region crop → feature extraction → feature vector. The package specifies the person class ID, preprocessing and feature vector format.
Read the Body SDK overviewIntegration materials
Version-specific API headers, sample code, model files and compatibility notes are supplied after review. Public downloads are not available for commercial SDKs or model packages.
Request the evaluation materials