**** BEGIN LOGGING AT Fri Mar 06 03:00:04 2020
Mar 06 23:57:42 <pradan[m]> The current test have been made on the darknet model (YOLO model precisely) using NNpack for faster computation. The Caffe models have NNpack enabled by default, as mentioned in the docs. So, can we know whether the Caffe-Jacinto models present as pre-trained models by TI have that functionality too ?
Mar 07 00:03:59 <pradan[m]> I looked for other ideas.... and found an interesting one.
Mar 07 00:04:00 <pradan[m]> We could use FeatherCNN for further boost. It outperforms NNpack by 12 times, and gives state of the art inference results. The limitation being, they only support Caffe right now. So, we could either begin with Caffe _or _convert our darknet model to Caffe if darknet _is_ to be used compulsorily.
**** ENDING LOGGING AT Sat Mar 07 03:03:04 2020