Here are some tips for settings to improve the accuracy of DeepSky! (Annotation Edition)
This time, I will introduce information regarding annotations. Annotations are one of the important settings related to detection accuracy, so I hope you find this useful. Q. Should labels be grouped together or divided into finer categories? When there are multiple detection targets (such as scratches, dirt, and dents), there are two patterns: registering all annotations under the same label "defect" or dividing the labels by shape as "scratches, dirt, dents." DeepSky tends to show improved detection results with fewer labels, so it is generally better to register them under the same label. However, if it is necessary to know which defect has been detected, it is essential to separate the labels by shape. In this case, if a scratch is mistakenly registered as dirt, it can lead to inconsistencies during training, resulting in poor learning outcomes. Therefore, annotations must be carried out carefully to avoid mistakes and oversights. (Figure 1)
Inquire About This Product
basic information
Q. When different parts of different colors are grouped together, can the DS learn that "the shape of the parts should not matter and should be judged solely by color"? The same can be said for color rather than shape. When there are parts of different shapes and different colors, it is possible to teach the system to ignore the shape and learn only the color. Even with different shapes, by registering them under the same label, such as red for red and blue for blue, it is possible to teach that "the shape of the parts should not matter and should be judged solely by color." However, if it is necessary to detect them in a more detailed manner, such as "red part A, red part B, blue part C, blue part D," it is necessary to register separate labels for each.
Price range
Delivery Time
Applications/Examples of results
For more details, please refer to the related links or feel free to contact us.
catalog（1）
Download All Catalogs
Company information
"Achieve significant effects" while "keeping costs down." It requires endless effort despite its simplicity. This is why SkyLogic believes it is valuable to continue pursuing "cost-effectiveness" through means such as technology and services. EasyInspector2/DeepSky: An appearance inspection software with a cumulative installation record of 2,500. Effect: We have developed a high-precision and lightweight AI model. It can connect five cameras simultaneously to one PC. Cost: To reduce the cost of learning setup, it is no-code, and simply deciding the inspection mode determines the necessary processing. Expansion and horizontal deployment are possible at low cost. EasyMonitoring2: A system that accurately reads and records meters through cameras and notifies of abnormalities, eliminating time-consuming and costly patrol monitoring. Effect: Not only meter reading but also the confirmation of processing tanks and oil leaks can be automated using AI. Cost: It can connect to inexpensive cameras, up to 100, with one PC. There are no monthly fees. In addition to these, we offer various products for "cost-effectiveness." Please let us know your company's challenges and themes.