IR model Results Really Bad

[derek-zr]

[Required] Your device (RaspberryPi3, LaptopPC, or other device name): LaptopPC

[Required] Your device's CPU architecture (armv7l, x86_64, or other architecture name): I7 8700K

[Required] Your OS (Raspbian, Ubuntu1604, or other os name): Windows Subsystem Linux（Ubuntu1804) with 2019R1 openvino

[Required] Overview of problems and questions:
Thanks for your great project. The IR model perform very bad, and different from the PB result. Here is a example:(Yellow is pb model)

I also see some related issues like #23. But I think it's not the bug of the code, because the boxes is not displaced, but also have many wrong predictions. I wonder if the conversion process of mine is wrong.
Here is the json:
[
{
"id": "TFYOLOV3",
"match_kind": "general",
"custom_attributes": {
"classes": 4,
"coords": 4,
"num": 9,
"mask": [3,4,5],
"jitter":0.3,
"ignore_thresh":0.7,
"truth_thresh":1,
"random":1,
"anchors":[10,13,16,30,33,23,30,61,62,45,59,119,116,90,156,198,373,326],
"entry_points": ["detector/yolo-v3/Reshape", "detector/yolo-v3/Reshape_4", "detector/yolo-v3/Reshape_8"]
}
}
]

[derek-zr]

The confidence threshold is same. Both are 0.5

[xiaozhishang]

You can update openvino version to 2019,and try it again

[derek-zr]

You can update openvino version to 2019,and try it again

It's already the 2019 version. 2019.1.094

[xiaozhishang]

You can update openvino version to 2019,and try it again

It's already the 2019 version. 2019.1.094

I am working on windows,usehttps://github.com/PINTO0309/OpenVINO-YoloV3/tree/master/cpp.

[derek-zr]

You can update openvino version to 2019,and try it again

It's already the 2019 version. 2019.1.094

I am working on windows,usehttps://github.com/PINTO0309/OpenVINO-YoloV3/tree/master/cpp.

Thanks. But I wonder if it's because the conversion of pb to xml and bin. Would you mind add my QQ
for more detailed discussion? QQ: 348217470