23.05.2024
AI-based part classification teaching
Based on customer feedback, a key benefit of the EYE+ smart control system is its systematic and efficient approach to teaching parts while ensuring ease of use. Each step in the teaching process serves a well-defined purpose.
Our current focus is on the candidate selection step, which aims to distinguish correct orientation from incorrect orientation. For this step, based on the candidates isolated in the previous step, the user must define 15 well oriented candidates and 15 badly oriented candidates. With this sample of 30 candidates, the AI algorithm is trained and able to keep only parts in the correct orientation, with incredible results. No need to define a complex and time-consuming training model. Thanks to AI, users can define the desired orientation in less than a minute.
Why use the overview classification mode?
Due to shadows or parts close to the frame of the Asycube feeder, parts can look different and it is difficult to classify these candidates. In order to understand the context of the parts, an overview mode is now available with the relase of EYE+ 4.3. Thanks to this overview image, users can select the desired image and classify the candidates directly from the complete image.
New images can be added and selected at any time to train the AI algorithm in post-production. If the orientation of the parts are difficult to distinguish, the user can manually place the parts in a known orientation and acquire a new image. This way, all the good orientations will be on the same image and it will be easier and faster to train the AI.
Benefits of the overview classification mode:
- Easily classify candidates by viewing them all at once on the platform
- Simply select the class (good or bad) and classify parts with a single click
- Benefit from an intuitive display for increased classification efficiency
- Refine the AI algorithm with new images and part configurations post-production
For more details, access the online EYE+ user manual