About inference speed

[auroua]

I am using the eval_forward_time.py test the inference speed.
But the inference speed is quite slow, I only get around 0.20.

Forward time per img (b=1): 0.014 (Mean: 0.020)
Forward time per img (b=1): 0.014 (Mean: 0.019)
Forward time per img (b=1): 0.014 (Mean: 0.019)
Forward time per img (b=1): 0.014 (Mean: 0.019)
Forward time per img (b=1): 0.036 (Mean: 0.020)
Forward time per img (b=1): 0.014 (Mean: 0.019)
Forward time per img (b=1): 0.031 (Mean: 0.020)
Forward time per img (b=1): 0.011 (Mean: 0.020)
Forward time per img (b=1): 0.011 (Mean: 0.019)
Forward time per img (b=1): 0.011 (Mean: 0.019)
Forward time per img (b=1): 0.037 (Mean: 0.020)
Forward time per img (b=1): 0.013 (Mean: 0.019)
Forward time per img (b=1): 0.037 (Mean: 0.020)
Forward time per img (b=1): 0.011 (Mean: 0.020)
Forward time per img (b=1): 0.011 (Mean: 0.019)
Forward time per img (b=1): 0.011 (Mean: 0.019)
Forward time per img (b=1): 0.014 (Mean: 0.019)
Forward time per img (b=1): 0.038 (Mean: 0.019)
Forward time per img (b=1): 0.014 (Mean: 0.019)
Forward time per img (b=1): 0.014 (Mean: 0.019)
Forward time per img (b=1): 0.037 (Mean: 0.020)
Forward time per img (b=1): 0.036 (Mean: 0.020)
Forward time per img (b=1): 0.013 (Mean: 0.020)
Forward time per img (b=1): 0.014 (Mean: 0.020)
Forward time per img (b=1): 0.014 (Mean: 0.020)
Forward time per img (b=1): 0.037 (Mean: 0.020)
Forward time per img (b=1): 0.012 (Mean: 0.020)
Forward time per img (b=1): 0.014 (Mean: 0.020)
Forward time per img (b=1): 0.014 (Mean: 0.020)
Forward time per img (b=1): 0.014 (Mean: 0.019)
Forward time per img (b=1): 0.037 (Mean: 0.020)
Forward time per img (b=1): 0.011 (Mean: 0.020)
Forward time per img (b=1): 0.014 (Mean: 0.020)
Forward time per img (b=1): 0.036 (Mean: 0.020)
Forward time per img (b=1): 0.014 (Mean: 0.020)
Forward time per img (b=1): 0.011 (Mean: 0.020)
Forward time per img (b=1): 0.038 (Mean: 0.020)
Forward time per img (b=1): 0.039 (Mean: 0.020)
Forward time per img (b=1): 0.014 (Mean: 0.020)
Forward time per img (b=1): 0.014 (Mean: 0.020)

Thegpu is nvidia rtx 2080 Ti.
I have set the inference on gpu.
parser.add_argument('--cpu', action='store_true', default=False)

[xiaoyufenfei]

Please refer to our paper, FPS: 70+, speed test script please refer to https://github.com/xiaoyufenfei/LEDNet/blob/master/test/eval_forward_time.py In addition, I may rebuild the project later, put some Lightweight segmentation network put into a project, thank you