TMDSCNCD28P55X: TMDSCNCD28P55X

Seungwoo Nam

Part Number: TMDSCNCD28P55X
Other Parts Discussed in Thread: TIDA-010955

Tool/software:

Hi, I was looking through this device's design guide and application in arc fault detection using embedded AI models.

Since this device is TIDA-010955 with control card C2000, I was wondering that if I can get information about which type of CNN is leveraged in C2000's Neural-Network Processing like 1D CNN or 2D CNN.. etc.

Thank you.

9 months ago

0 Yuan Zhao 9 months ago

TI__Expert 4055 points

Currently TI-NPU supported layers are documented here:
https://software-dl.ti.com/mctools/nnc/mcu/users_guide/layer_patterns.html
https://software-dl.ti.com/mctools/nnc/mcu/users_guide/layer_configs.html

We support both 1D and 2D CNN, 1D is usually represented as Hx1 (width being 1). We are expanding the layer configs that can be supported on TI NPU.

+1 Seungwoo Nam 9 months ago in reply to Yuan Zhao

Prodigy 60 points

Thank you for your reply.

I can now see that I can use my own trained model into this device.

Just one more thing to find out is that when training model with 'Edge AI Model Composer', there are 3 types of model I can select as shown below.

- ArcFault_model_200

- ArcFault_model_300

- ArcFault_model_700

In these models, I saw that they are based on CNN. Among them, which type of CNN is used? Like 1-D CNN or 2-D CNN as I mentioned earlier.

Thank you for your support.

0 Amir Hussain 9 months ago in reply to Seungwoo Nam

TI__Expert 5025 points

Hi Seungwoo

They are 1-D CNN.

Thank you

Amir hUssain

C2000™︎ microcontrollers

C2000 microcontrollers forum

TMDSCNCD28P55X: TMDSCNCD28P55X