A novel neural network architecture that can concurrently achieve real-time performance, competitive accuracy, and strong generalization ability.
307
stars
28
forks
source link
前辈，我在4090和3090上都进行了复现，在第二轮训练了30个epoch的情况下，epe还是只收敛到了0.67，其他指标也都比论文中稍差，请问这是什么原因呀？谢谢前辈！ #19
Closed
flyxuexi closed 3 months ago