the inference time cost two much?

[longzeyilang]

I use a picture to detect, the model is Decoupled_SOLO_Light_R50_3x and the default config. the inference time cost two much time,about 0.6 sec,why?

[WXinlong]

@longzeyilang Can you provide your script used to test the speed?

[longzeyilang]

use default config. And nvidia is 1080ti

[WXinlong]

@longzeyilang I mean how you test the speed. It looks like you test through the demo script, which includes additional initialization and processing steps like model loading and visualization part.

[longzeyilang]

https://github.com/WXinlong/SOLO/blob/master/demo/inference_demo.py

[WXinlong]

@longzeyilang The inference speed should be tested in test_ins.py, not the inference_demo.py, due to the reasons stated above.

[longzeyilang]

what is difference?

[WXinlong]

@longzeyilang The inference_demo.py includes additional initialization and processing steps like model loading and visualization part.