Tool/software: TI-RTOS
Hi,
Could you please confirm whether i understand the process of TIDL OD usecase correctly?
I refered the source code and flow diagram of TIDL OD.
1. IPU get the input frames and decode / Video processing / Duplicate
2. A15 convert RGB format frames to YUV format
3. 4 EVE cores do computation of layers that allocated to EVE cores (by using LayerGroupsID)
4. DSP also do computation of layers that allocated to DSP
5. IPU merge the input frames and result of DL network
6. TIOP on DSP
7. IPU draws objects
8. Display the result
In addition, i also want to know what is the role of Merge and Object propagation.
Actually, i did read TIObjectPropogation_DSP_UserGuide but can not understand easily.