Team Project:
Huihe – “Cloud + Client” Intelligent Visual Recognition Solutions

07.2017-12.2018

Heils Technology Co., Ltd, Shenzhen

Shenzhen, Guangdong, China; Xi’an, Shaanxi, China

Heils Technology Co., Ltd, Shenzhen (深圳禾思众成科技有限公司) is an AI startup company focusing on computer vision, with a core team formed by students from CMU, UIUC, NUS, XJTU and other top universities around the world.

Founded in 2015, the company focuses on deep learning computer vision algorithms, computing platforms and big data application research, where I was involved in their team providing leading Huihe – “Cloud + Client” (慧禾“云＋端”) intelligent vision solutions with collaboration of labs and industries.

Heils “Cloud + Client” solution

The “Cloud + Client” intelligent visual solution is a Computer Vision based commercial solution with combination of hardware and software, in which the cloud part focuses on structured data storage and data mining, including remote deep learning neural networks and other machine learning models; the client part consists of web, mobile phones, PCs, cameras and other devices, performing image capture and recognition functions.

Heils “Cloud + Client” workflow

The algorithms are deployed on both the front and back ends, enabling intelligence on the “Cloud + Client” at the same time. As the front-end device being portable, lightweight and compact, this solution costs less while remaining competitive when compared to traditional CV solutions that imports specially designed devices and could easily cost thousands of dollars, making AI more accessible.

Heils “Cloud + Client” intelligent visual recognition system 

The solution was tested and implemented in 2 major domains: OCR in complex scenarios and AI-enabled traditional industries. For the complex scenarios including dim light, backlighting, glare, characters worn, partially hidden, partially obscured, etc., the system could adaptively recognize numbers, letters, Chinese characters and mathematic formulas, with achieved recognition rate reaching advanced level in the commercial use.