Pascal

Results 3 issues of Pascal

0.0.4版本第26页处,您写道: > 其实,在创建网络实例的过程中, 一旦调用 nn.Conv2d 的时候就会有对权值进行初始化 Conv2d 是继承_ConvNd,初始化赋值是在_ConvNd当中,这些值是创建一个 Tensor 时得到的,是一些很小的随机数。 实际上,并非如此,在pytorch源码中使用的是kaiming初始化,也就是说,pytorch的模型权值初始化不是很小的随机数。 参见: https://github.com/pytorch/pytorch/blob/6e453e56f991d43ff9d0eac715020b7ef877ca77/torch/nn/modules/conv.py#L43 https://github.com/pytorch/pytorch/blob/6e453e56f991d43ff9d0eac715020b7ef877ca77/torch/nn/modules/conv.py#L45

https://github.com/milongo/CASENet/blob/f84246f200b12335551f06a34c840f790401dba8/models/resnet.py#L118 kernel_sz here should be 1, according to the paper. https://github.com/milongo/CASENet/blob/f84246f200b12335551f06a34c840f790401dba8/models/resnet.py#L153 Here is wrong, your method seems like DSN, not CASENet. https://github.com/milongo/CASENet/blob/f84246f200b12335551f06a34c840f790401dba8/models/resnet.py#L195 I think you should use: acts = self._fused_class(sliced_cat,...