TDA4AEN-Q1: NPU Benchmarks for TDA4AEN

Reposting the table that Alan linked above as it came across as blurry for me:

Hi Alan,

Can you please explain what you mean by NPU?

Regards,

Chris

NPU = Neural Processing Unit, I guess TI refers to that as the Deep Learning Accelerators

Hi Alan,

Please take a look at these pages.  They have general overview of performance.  The stats are a bit older (about 1 year or so), but still relevant. 

https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-jacinto7/06_01_01_12/exports/docs/tidl_j7_01_00_01_00/ti_dl/docs/user_guide_html/md_tidl_accuracy.html

https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-jacinto7/06_01_01_12/exports/docs/tidl_j7_01_00_01_00/ti_dl/docs/user_guide_html/md_tidl_performance.html 

Regards,

Chris

The performance data you provided does not cover the benchmarks we need.  Could TI please measure performance with the benchmarks we requested?  

If we were to prioritize it would be ResNet50, Mobilenetv3, EfficientNetv2 as highest priority.