• State
  • Locked Locked
  • Replies 1 reply
  • Subscribers 31 subscribers
  • Views 90 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

MSPM0G5187: how to improve char recognition example?

Frank(EP) Liu
Expert 5260 points

Part Number: MSPM0G5187

hi,

customer uses char recognition example and input "A" "B" "C". but it still shows high chance to 0-9.

They don't change any thing from example below. Can I know how to improve it?

https://dev.ti.com/tirex/explore/node?isTheia=false&node=A__AHBmqPnl3CbtUp5N8nxXTw__MSPM0-SDK__a3PaaoK__LATEST

can I add 11th classes and put "A" "B" "C"... not 0-9 to this classes?

 

BR,

frank

 

  • 0
    TI__Genius 12315 points

    Hi Frank,

    I believe the CNN module used in this example is trained by the characters from '0' to '9', so the application will also focus on the recognition of those characters.

    It is more an example to show the use case of NPU module in MSPM0G5187, and if customer wants to add more recognition characters such as "A", "B", "C", they may need to training there own CNN module with training set data including those characters. 

    The guidance of training customer's own module could be found in Edge AI User Guide.