AM69A: For Vision system

Part Number: AM69A
Other Parts Discussed in Thread: TDA4VM

Tool/software:

Hi

We are planning to use AM69A for our vision system which would be used for detecting PCB fault during assembly.The vision system should be capable of handling mass volume, and be fast and accurate. Please let us know if AM69A would be a good fit.We initially planned with TDA4VM but considering the AI capabilities of AM69A, we are now planning to use this. Please let us know if this is a good choice.

The Vision system is required to detect fault like missing components(0201 package also),misaligned components,open/shorts,identify scratches,all soldering defects.

  • Hello Prema,

    Thank you for considering the AM69A for your PCB assembly defect detection application. 

    TIDL is able to support a variety of use-cases, and in your case, my recommendation would be an Object Detection Model. Our ModelZoo has a variety of different model that can be used as a backbone to your model to be able to do these tasks. This will take some modifications and model training on your end to get the model accuracy high for the purpose you are trying to use it for, then deploy on the hardware. 

    Model Zoo: https://github.com/TexasInstruments/edgeai-tensorlab/tree/main/edgeai-modelzoo/models/vision/detection

    Someone from our marketing team will respond to assist with which device would be best for your needs. 

    Warm regards,

    Christina

  • Hello Prema, 

    It is nice to meet you. As Christina said, the Object Detection Model would be the best path to follow with what you are trying to accomplish. As for selecting the correct device, please refer to our product pages on TI.com to see the features and offerings that best match with your application needs. AM69A is a device that offers the maximum number of TOPS and high-performance compute. We have scalable solutions that would also fit your needs if you decide you do not need something with that much AI capabilities. 

    Another resource I would recommend, would be our Edge AI Studio. In here, you can bring your own data set to test against our models to see which device fits best. 

    Do you have a specific TOPS number that you are trying to target?