当前位置:网站首页>Sophon autocv: help AI industrial production and realize visual intelligent perception
Sophon autocv: help AI industrial production and realize visual intelligent perception
2022-07-05 17:57:00 【Star Ring Technology】
Perceptual intelligence maps signals from the physical world to the digital world , yes AI The only way for industrialized production to come to fruition , Among them, visual perception and IOT perception have become the technical cornerstone in the field of industrial Internet of things , Through the combination with edge calculation , Can effectively solve AI Massive data processing and real-time response in the process of landing 、 Low value density of raw data 、 Discrete processing of multimodal data .
Sense the direction of intelligent business
The main process of realizing edge intelligent perception , With CV From the perspective of application landing , Mainly around model production and deployment . A typical CV The process required for application landing is long , Need experience “12 Stage ”, And it requires a lot of manpower , from 5 Intermediate engineers and 5 Senior engineers jointly complete .
The present tradition CV Pain points of application : Overall process leader , There are many people involved , The nonstandard process links and the fragmentation of the whole process will lead to the fragmentation of model management 、 Dealing with change takes time and effort 、 The last mile of model and application 、 Difficulties in continuous operation of the model . thus , Need an overlay data processing 、 Model training and a platform supporting business customization functions to realize the integrated production application and management of models .
Tradition CV Application business pain points
Against tradition CV Application business pain points , Star Ring Technology launched a one-stop model production application platform ——Sophon AutoCV.Sophon AutoCV Facing the inside of the enterprise CV Modeling team , Provide guided model training and the ability to build and deploy low code applications , solve CV The process of model production and application is long 、 Low efficiency , To shorten the CV Model production deployment cycle , Realize efficient operation of business applications .
AutoCV As an intelligent platform for visual perception , Change the business process from traditional “12 Stage ” Reduced to “ The underlying resources - Model training and management - Model application - Structured data return ”, Comprehensively cover the data in the process of model landing 、 Model 、 application 、 Management requirements for iterative operations . In addition to this ,AutoCV Focus on tradition CV Pain points of application : There are many people involved , From required “5+5 intermediate / Senior engineer ” Reduced to “2+2+2 senior / intermediate / Junior Engineer ”.
Business support cases
Multimodal data fusion processing
A process manufacturer , Based on this platform, the unified access of image and sensor data and model application are realized , Realize intelligent auxiliary production .
Business pain points
1) Manual operation process error : The manual feeding sequence affects the production quality , However, workers often fail to find the prompt in time when they make wrong feeding ;
2) The quality removal effect is not ideal : In the production process , There is much room for improvement in the identification and removal effect of impurity removal .
Solution
1) Manual feeding sequence monitoring prompt : Identify the material packaging through the on-site monitoring camera , Then identify the feeding sequence , The possible feeding sequence is wrong , Prompt through on-site audible and visual alarm ;
2) Image optimization algorithm of docking impurity remover : By docking the high-speed industrial camera in the trash remover , Collect impurity samples for marking , General model of continuous training iteration , Then optimize the different impurity removal instruments , Finally, the overall impurity removal effect is improved 2%.
Customer value
1) Manual feeding errors shall be timely captured and corrected , The feeding effect is significantly improved after operation .
2) The overall impurity removal effect in the production process is improved 2%, Reduce raw material consumption and product quality .
Enterprise empowerment cases
Process asset accumulation and model iteration
The Information Technology Department of a company , Realize the accumulation of multiple digital assets based on this platform 、 Model centralized scheduling and continuous iteration .
Business pain points
1) The model cannot be traced , The original data storage is disordered , There is no unified management method for various business data ;
2) Hardware resources in decentralized business systems cannot be centrally managed and scheduled ;
3) The algorithm model changes with the business , Accuracy can no longer meet the needs of the business , A single point of optimization is required .
Solution
1) Platform unified asset accumulation : The platform supports model training sample management , Support model multi version evaluation and management ; Support the accumulation, migration and reuse of business rules ;
2) Centralized scheduling of hardware resources : Rapid deployment of models in customer business systems , Share hardware resources at runtime ;
3) The algorithm model continues to iterate : Support the synchronization of structured data after model processing to the training module , The rapid iteration of the model can be completed through incremental training .
Customer value
1) data 、 Model and application asset accumulation , Value precipitation ;
2) The utilization of hardware resources is improved 20%;
3) Spiral iteration , Model iteration cycle is shortened 50%.
Star ring technology is fully self-developed Sophon AutoCV At present, it has shown strong performance and high stability in the practice of many industries . Overall speaking , Star Ring Sophon AutoCV Its core advantages can be summarized as the following five points :
- Efficient production : Through intelligent pre bid , Tagging is more efficient than open source tools promote 30%, Platform operation based on the whole process , The manpower input for model production can be Reduce 50%;
- Zero threshold operation : The platform supports model production and deployment in a guided and low code manner , Business operators You can also get started quickly , And complete model iteration and optimization ;
- High performance operation : The platform supports reasoning cards in addition to the mainstream , And support Localization Such as huawei 、 The model operation of Cambrian and other hardware is accelerated , To support Ten thousand way camera Intelligent stream processing ;
- Multi asset accumulation : For enterprises in the process of digital intelligence transformation 3 Class digital assets , data 、 Models and applications , It can effectively accumulate and precipitate through the platform , Reflect the vision AI Team value ;
- Open cooperation : The platform can cooperate with different partners , Up to now, it has exceeded 20+ We will cooperate with our friends to deliver the overall scheme , These include colleges and universities 、 Algorithm provider 、 Channel, etc .
We believe that , as time goes on , Star Ring Technology Sophon AutoCV It will continue to empower thousands of families , Help vision AI Industrial production , And vision AI A burst of value .
边栏推荐
- 论文阅读_中文NLP_LTP
- 提高应用程序性能的7个DevOps实践
- Humi analysis: the integrated application of industrial Internet identity analysis and enterprise information system
- Data access - entityframework integration
- Cmake tutorial Step2 (add Library)
- 2022新版PMP考试有哪些变化?
- matlab内建函数怎么不同颜色,matlab分段函数不同颜色绘图
- Thesis reading_ Chinese NLP_ LTP
- Sentinel flow guard
- "Xiaodeng in operation and maintenance" is a single sign on solution for cloud applications
猜你喜欢
随机推荐
Server configuration jupyter environment
基于YOLOv3的口罩佩戴检测
提高應用程序性能的7個DevOps實踐
Cmake tutorial Step2 (add Library)
小白入门NAS—快速搭建私有云教程系列(一)[通俗易懂]
Leetcode daily question: merge two ordered arrays
企业数字化发展中的六个安全陋习,每一个都很危险!
Easynmon Usage Summary
Troubleshooting - about clip not found Visual Studio
SQL Server(2)
毫无章法系列
rsync
Compared with the loss of Wenxin, the performance is improved a lot
从类生成XML架构
Oracle recovery tools -- Oracle database recovery tool
Neural network self cognition model
读libco保存恢复现场汇编代码
Please tell me why some tables can find data by writing SQL, but they can't be found in the data map, and the table structure can't be found
QT控制台打印输出
Ten capabilities that cyber threat analysts should have