1. Built-in self-developed 4th generation NPU, up to 0.5TOPs computing power
RV1106 and RV1103 adopt Cortex-A7 CPU and high-performance MCU, built-in fourth-generation NPU self-developed by Rockchip, which has high operation precision and supports mixed quantization of int4, in8 and int16, in which int8 computing power is 0.5TOPs and int4 computing power can reach 1.0TOPs.
2. Built-in self-developed third-generation ISP3.2 to support a variety of image processing technologies
RV1106 and RV1103 adopt the third generation ISP3.2 developed by Rockchip, which supports 5 million and 4 million pixels respectively, and supports various image enhancement and correction algorithms such as HDR, WDR, multi-stage noise reduction, and so on. In all kinds of complex light scenes, the effect of black light full color and backlight strong light shooting is visible.
RV1106 and RV1103 can support 2-3 MIPI/DVP inputs, which is the preferred scheme for economical binocular vision products.
3. Strong coding ability, high frame rate, low bit rate, small footprint
In terms of video encoding, Rockchip RV1106, and RV1103 have super encoding performance, support intelligent encoding, and save code stream according to the scene adaptively, saving more than 50% of the code rate compared with the conventional CBR mode, so that the shooting picture is both high-definition and small in size, double the storage space. It also supports rich encoding functions, such as frame skip reference, custom quantization matrix, subjective factors, etc., to further improve the encoding quality.
4. Intelligent audio and sound recording are clearer
In terms of audio processing, Rockchip RV1106, and RV1103 adopt intelligent audio solutions, support echo cancellation, voice noise reduction, cry detection, abnormal sound detection, etc., support high-definition voice, enhance the sound collection, and long-distance sound pickup.
5. Fast startup transient response, high performance, and low power consumption
RV1106 and RV1103 are designed with a built-in RISC-V MCU, which supports low-power fast startup, supports 250ms fast snapshot, and loads the AI model library at the same time, which can realize face recognition "within 1 second".
6. High integration
RV1106 and RV1103 have the built-in Audio codec, MAC PHY, RTC, etc., and provide the QFN package with built-in DDR and BGA package without built-in DDR.
CPU |
RV1106, Cortex A7 + MCU |
RAM |
Built in 1Gb~2Gb DDR3L |
ROM |
2Gb NAND Flash,Support NOR Flash |
NPU |
RV1106 has 0.5TOPS, Support RKNN AI framework, which can realize the conversion of common AI framework models such as (Caffe, darknet, Mxnet, ONYX, PyTorch, TensorFlow, TFlite) and algorithm |
Sensor |
5 million pixels SC530AI Sensor,Monocular camera |
CSI |
4 Lane MIPI CSI,Support 500 million pixels@25fps,Support up to 3 sensor access |
DSI |
NA |
WIFI |
SDIO WIFI,IEEE 802.11b/g/n |
4G |
USB 4G module,Support CAT4,CAT1 main stream modules |
Net port |
Adaptive 10/100Mbps 100M, Support MDIX function |
USB |
OTG2.0 X1 |
MIC |
Analog Omnidirectional MIC |
SPK |
Independent external Audio Coder, pickup and Speak, with 3W power amplifier, no Support Line out |
Photosensitive |
Optional (ISP night sensing function). |
Fill light |
Expandable LED fill light panel, support white light/IR fill light |
Reset |
The port does not elicit a reset signal |
ISP |
Third generation ISP,5M30 2F HDR/3NDR/WDR/BLC/DPCC/PDAF/LSC |
Coder |
H.264/265 5M30FPS,5M@60FPS JPEG snapshot,Six bit rate control modes (CBR, VBR, FIXQP, AVBR, QPMAP, and CVBR) |
System support |
Linux (Support fast start, first valid frame rendering time < 150ms). |
Operating temperature |
-20℃~70℃ |
Operating platform |
MCU、ARM、Windows、Linux、Android |