"Paddle + camera" has become a "prefabricated dish" in the AI world, and it is easier to implement industrial AI quality inspection

This article has been published on the official account of the flying oar , Please check the link ：
“ Flying propeller + Photographic camera ” Become AI Bounded “ Prefabricated dishes ”, Industry AI It is easier to implement quality inspection

once , He is known as “ Computer prodigy ”、“ Rock boy ”, He first developed “ Human skin grinding ” Software once occupied 80% of the market segment . Now , In his middle age, he went with the tide and became a Tiktok blogger , At the same time, it continues to play the role of yard farmer in the market “ killing ”, Founded the shadow recognition technology .

Founder of discerning technology Wang Jingjing

Wangjingjing, founder of discerning technology, said ,“ From small to large , I write code rather than composition 、 More than talking ”. This time, , He hopes that his technical ability can make some contributions to the intelligent upgrading of traditional manufacturing industry .

Intelligent manufacturing is feasible , But it's difficult

“ the 14th Five-Year Plan ” Proposed in the intelligent manufacturing development plan , To 2025 The main objectives of the year include ,70% Manufacturing enterprises above Designated Size have basically realized digitization and networking , Production efficiency of manufacturing enterprises 、 Product yield 、 The utilization rate of energy and resources has been significantly improved , The level of intelligent manufacturing capability has been significantly improved .

For the development of manufacturing enterprises , The production efficiency and yield of products are very important , And quality inspection work and production efficiency 、 The yield is closely related , Only qualified products can enter the market . In traditional manufacturing enterprises , The quality of products is tested by manual or traditional visual algorithm . The manual quality inspection method has different inspection standards 、 The cost of personnel training is high 、 Problems such as false detection and missing detection , Directly affect the product quality inspection results . The cost of traditional visual algorithm detection is high 、 The effect is often not as expected , The production line is still inseparable from high strength 、 Highly repetitive human participation .

“AI Technology has many advantages , But enterprises can really apply AI Solve practical problems , There are still many difficulties to overcome .” Wangjingjing found , The visual recognition ability of artificial intelligence based on deep learning can indeed reach the level of excellent artificial quality inspectors , But hire according to your own needs AI The cost of Algorithm Engineers' development and maintenance is relatively high , Not every enterprise can afford ; On the other hand , At present, the domestic excellent AI Algorithm engineers are scarce , Even fewer can focus on manufacturing ; Besides , Deployment hardware for industrial scenarios Also is AI Application landing “ Last mile ” problem .

Facing these problems, which lie on the road of intelligent upgrading of manufacturing industry “ stumbling block ”, Wang Jingjing believes , Low threshold 、 One stop integration of hardware and software AI Solution , Let data collection and reasoning deploy 、 The model iteration is completed on a special camera for industrial production line , It is an efficient path that can be landed .

Flying propeller EasyDL+ Photographic camera

Give Way AI Quality inspection is simple and efficient

solve AI The problem of technology landing in the industrial production scene , Wang Jingjing has gone from soft 、 Hardware has found the answer .

Enterprise version EasyDL Let wangjingjing see the zero threshold AI The development platform is fast 、 convenient 、 The power of efficiency , He has repeatedly used figurative metaphor to express his feelings about the use of this platform ,“EasyDL Modularize many algorithms ,‘ Just like a chef doesn't have to make a kitchen knife himself ’, Coders can call on demand , Development AI Program 、 To make a AI Hardware , Whether it's PC、 Flat 、 TV or Android 、iOS, Can be adapted to .”“ If we use other AI Developing architecture and modeling is like climbing stairs , That's it EasyDL It feels like taking the elevator , To save time 、 Save effort and worry .”2020 year , Wang Jingjing became one of the first batch of propeller developers and technical experts certified by Baidu PPDE.

Hardware support , Developed by discerning film technology AI QC edge computing camera （ hereinafter referred to as “ Photographic camera ”） Simple deployment 、 Interface is rich , With high sensitive LCD touch screen , The model can be managed without external display , And the shadow discrimination can be connected with a buzzer , Give a prompt to the effect verification . The model trained through the propeller platform can be easily deployed to “ Shadow discrimination ” On , The whole process doesn't need to write any code , It's solved AI The last mile of algorithm deployment .

at present , By flying oars EasyDL And a photographic camera AI Quality inspection combination has been applied to Engine leakage detection 、 Piston ring defect detection 、 Weld bubble detection 、 Rivet assembly inspection 、 Injection molded parts are not fully filled and are missing 、 Inspection of thread opening defects Waiting for work . Take the flaw detection of thread port as an example , First, the data is collected according to the sample characteristics , And then the oars EasyDL Mark and train the qualified and unqualified samples on the platform , Finally, it is deployed to the photographic camera . In the production line, if the image recognition camera detects that the parts are unqualified, it will beep or be removed .

“ We just need to meet the needs of the factory quality control department , Flying oars EasyDL Mark the data of good and defective products on , The model can be trained completely according to the standards of the factory quality inspection department , Data annotation and model training do not require QC personnel to understand the code .” Wang Jingjing believes , It's like a shop that sells prefabricated vegetables without the presence of a chef , Photographic camera + Enterprise version EasyDL Namely AI Bounded “ Prefabricated dishes ”, No, AI Algorithm engineers can also be easily applied in the factory AI, Authors efficiency .

Wang Jingjing said , real AI It should be very simple , I hope you won't be scared away by technical terms , In fact, everyone can pass “ Flying propeller + Photographic camera ” Realize the model effect developed by Algorithm Engineers , Realize AI Application changes from difficult to easy , Realize in your own field AI Innovation breakthrough .

Read more

How to conduct industrial quality inspection with “ wisdom ” To win ？ The whole process scheme of industrial parts testing

Industrial parts AI Detailed explanation of visual quality inspection scheme , The whole process scheme of bottle cap defect quality inspection

Focus on 【 Flying propeller PaddlePaddle】 official account
Get more technical content ~