I had a try according to this “https://docs.openvinotoolkit.org/latest/_inference_engine_tools_calibration_tool_README.html”
and had changed the fp32 model to int8 model by the calibration tool in the simplified mode. but the accuracy loss is too large.
So I start to test the model calibration flow, but do not know how to Convert data annotation files. the example is a classification
case,the labels are words, and in the semantic segmentation the labels are pictures, i can not passing parameters in the same way as the example
Hemanth Kumar G. (Intel) wrote:I had a try and find that it is so difficult to convert a PF32 model to INT8，because the command-lines have many difference between semantic segmentation case and classification case .Could you send me some examples or any detailed parameter descriptions？thank you
You can try the solution provided in the following discussion - https://software.intel.com/en-us/forums/computer-vision/topic/810212
Dear liang, heng,
DL Workbench is an early tool (in preview model) and still doesn't have full functionality. So you can always "go stick shift" (meaning use the manual method) using our Calibration tools. This github issue 171 post which I replied to is slightly old but it's still valid. If DL Workbench doesn't work for you the manual tools will.
Hope it helps,