BUSINESS

EZ Defect

AI Defect Image Generator
EZ Defect is software that uses a deep learning algorithm to create and provide a fake image similar to defects generated in the industrial field. When introducing AI to industrial fields, the class imbalance problem, in which the quantity of defect data is insufficient compared to normal data, is a chronic problem found in every industrial field. LAONPEOPLE’s EZ Defect provides a solution that users to create a defect image similar to the real one. Users can create defects in desired shape and size to solve the class imbalance problem.

Why LAON PEOPLE
1. Fast and Stable
- The time of AI introduction can be accelerated by using the fake Images generated through EZ Defect as the initial learning data. In addition, early obtained defect data helps to stable the performance of AI.
2. The last 1%
- Even if defect data was obtained, performance may stagnate at certain points in AI training. Ez Defect allow user to create defect in desired location and shape quickly and easily, those defect images generated by Ez Defect can help to solve the stagnation of AI performance.
3. More easily
- EZ Defect learns with small data easily. You can generate the defect data you want through a user-friendly UI.
INQUIRY
sales@laonpeople.com031-707-7052ONLINE INQUIRY
LIST VIDEO

Product Specific Specifications

Information
SCROLL >>
Not registered post.
EZ Defect Information

 




Introduced AI?

How is building your data set going on?

The most difficult thing to face when introducing AI in the industrial field,
is to build data sets.

 

Small data sets
In the early stages of AI introduction, it is difficult to extract inspection performance due to not enough data sets for learning yet.
Short building time
It takes a long time to obtain sufficient data, but in real industry field, most projects need to be built within a short period of time.”
Data imbalance
Common defects gathered too much, and intermittent defects are very difficult to gather.
No matter how good AI model is, if it’s not supported by
a great data setit will fail to achieve the expected effect eventually.


Then, how to make a great data set?




※ What’s Good Data?
Refined data that can effectively help AI learning by minimizing
the noise inherent in the data and representing each data type.
 

Insufficient data
The whole learning process
is governed by the noise.		 Using Big Data
The effect of noise is minimized, but data collection, refinement, and learning time can be delayed. (inefficient)		 Using Good Data
With good data with minimized noise, excellent results can be obtained even with a small amount of data.


Use Good Data

No matter how good an AI model is used in the industrial field,
in order to learn AI model efficiently, we need good data rather than big data

Everyone knows it's good to use Big Data, but it takes too much time and resources
to gather a lot of data and to refine and evaluate in the industrial field





Then, how to make Good Data?

EZ Defect is the answer!

EZ Defect is a powerful image generation solution provided by LAONPEOPLE,
and it is a solution that helps to obtain insufficient defect data effectively by generating good data
which is close to the defects that actually occur in industrial field.

The 'GAN' algorithm is used to generate natural defect images in desired location and shape.
With EZ Defect, you can not only accelerate the introduction of AI machine vision system by obtaining defect data,
but also achieve dramatic performance improvements!






Generation and learning time (with RTX 3080 GPU)
- Learning: 1hr to 1hr 30 mins. (approx.)
- Image Generation: 5~6 secs (in 1024 x 1024 resolution)



Faster, with a single Image!

EZ Defect is a solution that can be used in various industrial domains such as semiconductors,
electrical and electronics, textiles, metals, and pharmaceuticals. To learn rare defects, you can proceed learn with a single image.
EZ Defect Tool is used to label the defect location provided with user-friendly interface.


Every spot you want!

Using the EZ Defect Tool, select the type of defect that you want to generate,
and label the defective area on the image,
then defect image is created with the original image resolution!


Will the images you created really help learning?

EZ Defect generates images with minimal damage to the original image.
Therefore, the image generated through EZ Defect also has almost the same value as the original data,
this means images generated by EZ defect can be used as effective learning data.



The inspection accuracy can be improved significantly by obtaining rare defect data.



Take advantage of EZ Defect!

EZ Defect can solve various problems that occur in industrial fields.




 Bad data acquisition problem
Although it takes a long time to collect intermittent bad data,
EZ Defect can effectively shorten the time it takes to introduce real AI by creating the defect image you want.

Data imbalance problem
No matter how good the AI performance is, if there is an imbalance of
data between good/defective or defective types, it will not perform properly.
EZ Defect can solve those data imbalance problem and this will ensure data diversity also.

AI model performance issues
AI models can be vulnerable to certain types of defect.
If you create weak types of defect through EZ Defect and use it as training images,
you can improve the accuracy of the AI model.