为什么mmsegmentation中的deeplabv3+模型在aspp之后,代码里用的是33卷积核?不应该是和论文里一样是11的卷积核吗?
这要看你怎么设置了,你选用哪个模型,用的就是哪个
以下答案引用自GPT-3大模型,请合理使用:
片段。
The number of convolutional layers and kernels were tuned during the training process. Specifically, a deep Convolutional Neural Network (CNN) with 11 layers and 330 kernels was found to give the best performance on the Stanford Dogs dataset.
如果我的回答解决了您的问题,请采纳我的回答