Hi3519A V100
Hi3519A V100 4K Smart IP Camera SoC
Issue 02 (2018-06-20)
HiSilicon Proprietary and Confidential
Copyright © HiSilicon Technologies Co., Ltd.
2
Introduction
Hi3519A V100 is a high-performance and low-power 4K
Smart IP Camera SoC designed for IP cameras, action cameras,
panoramic cameras, rear view mirrors, and UAVs. Hi3519A
V100 introduces H.265/H.264 encoding and decoding, with
performance up to 4K x 2K@60 fps and 1080p@240 fps.
Integrating with the HiSilicon fourth-generation ISP, Hi3519A
V100 provides WDR, multi-level NR, 6DoF DIS, and multiple
image enhancement and correction algorithms, ensuring
professional image quality. Hi3519A V100 also supports 4K
raw data output, facilitating post editing. With the advanced
low-power process and architecture design, Hi3519A V100
provides users with long-lasting battery life.
Hi3519A V100 integrates a powerful programmable
neural network inference engine and a vector DSP, backing the
application of multiple intelligent algorithms.
Hi3519A V100 supports multi-sensor inputs and
introduces a built-in high-performance AVS engine to
implement 4K 2-pipe to 4-pipe real-time panoramic video
stitching.
With the hardware-based 6DoF DIS, Hi3519A V100 has
reduced its dependence on the mechanical head during 4K@60
fps video recording.
Hi3519A V100 features efficient and ample computing
resources to assist customers in developing industry and
consumer applications. Backed by the dual-core A53 processor
and the DSP, Hi3519A V100 offers the dual-system solution,
enabling fast startup, real-time performance, and connections
with rich peripheral drives.
Hi3519A V100 uses the advanced 12 nm low-power
process and miniaturization package and supports
DDR4/LPDDR4 SDRAMs, allowing the product
miniaturization design.
With the stable and easy-to-use mobile SDK design
provided by HiSilicon, Hi3519A V100 can assist customers in
rapid product mass production.
Key Features
Low Power
1.9 W power consumption in a typical scenario for 4K x
2K (3840 x 2160)@30 fps H.265 encoding+neural network
algorithm
SVP
− A neural network inference engine (NNIE), a high-
performance DSP, and multiple CV acceleration
engines that facilitate customers' algorithm development
for achieving product differentiation
− Multiple neural network options
− 2.0 TOPS computing performance
− Complete APIs and toolchains
− Multiple applications such as face
detection/recognition and target detection/tracking
4K@60 fps Encoding
4K x 2K (3840 x 2160)@60 fps or 1080p@240 fps
H.265/H.264 encoding
Multi-Sensor Inputs
Up to 5-channel sensor inputs, supporting applications
such as panoramic camera and UAVs
Hardware-based Multi-Channel Video Stitching
2-channel 3K x 3K (3000 x 3000)@30 fps or 4-channel
1080p@30 fps stitching and recording
High-Speed Interface
PCIe, USB 3.0, and SDIO 3.0 interfaces that facilitate
customers' extension of external functional modules
评论3