search

594422814 / UDT_pytorch

MIT License
191 stars 23 forks source link

version problem #8

Closed swrdZWJ closed 4 years ago

swrdZWJ commented 5 years ago

Hi，Thanks for your excellent work，but I meet some difference from the original paper.Do you have a plan for a newer version and when do you release the complete version？Thanks ：）

594422814 commented 5 years ago

Hi, the results in our paper are obtained using the matconvnet version. Please refer to it for more details. Compared with the matconvnet version, the only difference is that this code omits the weighted loss, which only has a slight performance influence. Since I obtained similar results using this prototype, I did not further improve this code later. Sorry.