RuiZhang97 / ISNet

CVPR2022 ''ISNet: Shape Matters for Infrared Small Target Detection''
MIT License
124 stars 11 forks source link

Issue with reproducing the results #27

Open Lliu666 opened 9 months ago

Lliu666 commented 9 months ago

I followed the settings of 500 epochs as mentioned in the paper while running the open-source code, but it's challenging to achieve the results reported in the original paper. Additionally, I noticed that other researchers attempting to reproduce the results have also been unsuccessful in achieving the high performance mentioned in the paper. Could you please provide more details about the training process that may not have been explicitly mentioned in the paper?