[AI Image Inspection Case] Counting Processed Parts
We will inspect for any defective items while counting the number of necessary parts!
We often receive requests to count the number of necessary parts while inspecting for any defective items in inquiries to our company. Unmanned inspection and counting on a conveyor belt is a field in which DeepSky excels. This time, we received an inquiry about whether we could count processed parts (e.g., φ0.8×2.5mm) arranged randomly on a flat pallet. They provided images using backlight illumination. 【Inspection Settings and Results】 The left image shows a part of the work screen for the annotation process, where we outline the areas we want to identify (defects). As can be seen in the right image, all 10 annotated teacher images used were counted correctly. The ratio of successful counts for non-teacher images was 30%. There was a tendency for higher rates of miscounting with parts that were more closely stuck together. By using non-detected images for further training, we can detect defects that were previously undetectable. Further training is also one of DeepSky's strengths.
basic information
【Software Used】 Software: DeepSky Learning Edition Number of Inspection Points: Count the work present on the entire screen at 1 location.
Price range
Delivery Time
Applications/Examples of results
For more details, please refer to the related links or feel free to contact us.
