The TB-96AIoT is a low-power, high-powered core board for the AIoT field. It is equipped with a powerful neural network processing unit (NPU) and is compatible with a variety of mainstream inference models such as caffe and tensor flow. Together with the bottom board CarrierBoard developed, it can form a complete development board or evaluation board; the base board that can be customized according to the actual needs of the customer can directly from the industrial application board, which can meet the sweeping robot, drone, smart speaker. , automotive products, smart wear, security monitoring, AI computing modules and other areas of demand.

TB-96AI uses RK3399Pro as the main control chip, TB-96AIoT uses RK1808 as the main control chip.

Features:

Dual-core ARM Cortex-A53 CPU for ultra-low power consumption
Includes vfp v4 hardware that supports single and double precision operations.
ARM Neon Advanced SIMD (Single Instruction, Multiple Data) supports accelerated media and signal processing calculations.
Built-in neural network processor NPU, super high AI computing power
Compatible with Caffe/Mxnet/TensorFlow model, support multi-class framework, support mainstream layer type, easy to add custom layer
Provide easy-to-use development tools, PC-side model conversion, performance estimation, and accuracy verification
AI APPlication Develop Flow
Video decoder
Support H.264/AVC BASE/MAIN/[email protected]
Up to 1920×1080@60fp
Video encoder

*Support H.264 video encoder BP/MP/[email protected] level

*Up to 1920×1080@60fps

Camera interface

*MIPI-CSI×1, Built-in ISP image processor

*Maximum support input resolution 1920×1080

Display output interface

*MIPI-DSI×1

*Supports up to 1920×1080@60fps display output

Audio port

*Sperker×1

*Headphone×1

*Mic×1

*I2S×1

PCIE×1
USB2.0 HOST×1；
Type-C DRM×1；
I2C×3
LAN×1，1000M ETH
UART Debug×1
SPI×1
SD Card ×1
PWM×1
ADC×1, One for key input and the other for headphone insertion detection

Specifictions:

Basic Parameters
SoC	Rockchip RK1808(22nm FD-SOI)
CPU	Dual [email protected]
NPU	Support 8bit/16bit operation, computing power up to 3.0TOPS Support TensorFlow, Caffe model
VPU	1080p@60P H.264 Decoder 1080p@30P H.264 Encoder
RAM	1GB LPDDR3
Flash	16GB eMMC
Hardware Characteristics
Ethernet	Built-in Gigabit Ethernet PHY chip, 10/100/1000Mbps adaptive
WiFi/BT	Built-in WiFi/BT module, reserved antenna holder, can be directly inserted into the antenna
SoC	Rockchip RK1808(22nm FD-SOI)
Camera Interface	MIPI-CSI，Maximum support 1920 × 1080 resolution USB camera
Display Interface	One MIPI-DSI interface, up to 1920×1080@60fps display output
Audio Port	Sperker×1 Headphone×1 Mic×1 I2S×1
Type-C	USB3.0 DRM ×1
USB	USB2.0 HOST ×1
Extension Port	PCIE×1 I2C×3 SPI×2 SD Card ×1 PWM×1 ADC×2
Power input	DC 5V