How to create batch files for int 8 faster rcnn inference?


I’m trying to convert the sample faster rcnn model to int8 type. However, since I’m not quite familiar with Caffe, I don’t really know how to create the batch files for calibration. The tutorial in the developer guide is not very clear. I cannot even start from the first step.

patch -p1 < int8_caffe.patch

Where is this .patch file comes from?

Can anyone provide some detail steps of how to create this batch files?

I’ve downloaded a coco dataset for that, but really need to know how to convert them to the batch files. Thanks.