RK3399 has high performance, high expansion, all-round application features. CPU uses big.LITTLE size core architecture, dual-core Cortex-A72+ quad-core Cortex-A53, has technical leadership in overall performance and power consumption, GPU adopts quad-core ARM new The generation of high-end image processor Mali-T860, integrated with more bandwidth compression technology, the overall performance is excellent. Android 8.1 can fully call the neural network API (NNAPI) for hardware acceleration, so that RK3399's AI computing performance is greatly improved.
This release of the RK3399 SDK is a GPU acceleration solution based on Android 8.1 NNAPI, which provides accelerated support for AI-related applications developed with Android NNAPI. It has four advantages:
1. Wide compatibility: Standard API, directly supports various APK applications developed based on Android NNAPI;
2. Versatility: Supports many mainstream model architectures, suitable for various applications based on mainstream model architecture, including face recognition, ADAS, commodity identification, fatigue detection, etc.
3. Performance soaring: Real-time performance can be achieved in multiple tasks, such as using MobileNet for image recognition with a maximum frame rate of 23.2 frames;
4. Lower power consumption: Based on GPU efficient calculation, the full-load power consumption is only 1W;
According to the APK test data of image recognition and target detection, the performance of the mainstream model is excellent:
AI calculation is in the period of explosive growth. Rockchip artificial intelligence chip has been widely used in image recognition, intelligent security, intelligent driving, speech recognition, consumer electronics and other fields. This time based on the RK3399 platform Android 8.1 NNAPI optimization SDK, will make it The AI chip product line is more hard.