The RIKI machine vision chip is a UMC22nm process that supports 1T performance convolution operations and can accelerate deep learning inference computation for edge applications, including machine vision and machine auditory processing, to complete detection, recognition, and classification tasks. It has the characteristics of simple application, high integration, small size, and extremely low power consumption.
Contact ＞
Support INT8 quantization, optimize and improve computing power
Integrated high-performance ARM kernel, supporting secondary development
Support mainstream Caffe framework
Integrated image preprocessing capabilities, etc
|
characteristic |
|
|
Support INT8 quantization, optimize and improve computing power |
|
|
Support mainstream neural networks |
|
|
Efficient combination computing power, CPU dual core 2.0G paired with a maximum 1T professional neural network processing unit |
|
|
Can integrate 8Gb LPDDR3 |
|
|
Integrated image preprocessing capabilities, etc |
|
|
advantage |
|
|
High performance, low cost, low power consumption |
|
|
Strong universality: Corresponding to supporting all mainstream operators, covering common algorithm models and application scenarios |
|
|
Secondary development traversal: Advanced software architecture greatly facilitates user secondary development |
|
|
physical property |
|
|
structure size |
10mm×10mm |
|
Packaging form |
BGA256 |
|
Power supply voltage |
kernel：0.8V IO：1.2V、3.3V |
|
Environmental indicators |
|
|
working temperature |
-40℃ ~ +85℃ |
|
Storage temperature |
-55℃ ~ +125℃ |
|
Chip interface |
|
|
Functional interface |
2MB Sram |
|
LPDDR3, Up to 2133Mbps |
|
|
Supports SD/EMMC |
|
scan
