TangYong
TangYong
please tell me your email, i send all codes to you.
90 cord = torch.tensor(np.array([[i, j] for i in range(self.window_size) for j in range(self.window_size)])) --->
90 cord = torch.tensor(np.array([[i, j] for i in range(self.window_size) for j in range(self.window_size)])).type(torch.long)
to initialize session repeatedly lead to memory leak.
最近，我也开始继续研究分割，可以加我微信或者QQ：11355664，可以交流哈！
我是上班族，周末试试，再答复你哈，你自己也可以先研究！
modeling.py 71行， inplanes = 1024 #768, swin_t, swin_s, 1024, swin_b, 1536 swin_l
原因swin-transformer有不同的scale，导致特征输出层有不同的输出channel，如上面列举，你可以根据不用类型，设置不同的值。
inplanes = 1024 #768, swin_t, swin_s, 1024, swin_b, 1536 swin_l low_level_planes = 256 #192, swin_t, swin_s, 256, swin_b, 384 swin_l
不用类型或者scale的swin-transformer，特征长度（通道，channel）是不同的，就像我上面的答复那样！ swin_t和swin_s 底层特征通道数是768，高层特征是192 swin_b 底层特征通道数是1024，高层特征是256 swin_l 底层特征通道数是1536，高层特征是384 就是下面定义heads参数决定：24对应768和192，32对应1024和256，48对应1536和384 def swin_t(num_classes=1000, hidden_dim=96, layers=(2, 2, 6, 2), heads=(3, 6, 12, 24), img_size=224, **kwargs): return SwinTransformer(hidden_dim=hidden_dim, layers=layers, heads=heads, num_classes=num_classes, **kwargs) def swin_s(num_classes=1000, hidden_dim=96,...