lednet.patch

diff -x .git -x .gitignore -x '*.png' -r -N -r -u ./train/lednet.py esanet/src/models/external_code/LEDNet/train/lednet.py
--- ./train/lednet.py	2020-12-06 10:02:54.000000000 +0100
+++ esanet/src/models/external_code/LEDNet/train/lednet.py	2020-12-04 17:56:52.000000000 +0100
@@ -58,11 +58,11 @@
         x1 = self.pool(input)
         x2 = self.conv(input)
 
-        diffY = x2.size()[2] - x1.size()[2]
-        diffX = x2.size()[3] - x1.size()[3]
-
-        x1 = F.pad(x1, [diffX // 2, diffX - diffX // 2,
-                        diffY // 2, diffY - diffY // 2])
+        # diffY = x2.size()[2] - x1.size()[2]
+        # diffX = x2.size()[3] - x1.size()[3]
+        #
+        # x1 = F.pad(x1, [diffX // 2, diffX - diffX // 2,
+        #                 diffY // 2, diffY - diffY // 2])
 
         output = torch.cat([x2, x1], 1)
         output = self.bn(output)
@@ -205,7 +205,7 @@
         self.size = size
         self.mode = mode
     def forward(self,x):
-        x = self.interp(x,size=self.size,mode=self.mode,align_corners=True)
+        x = self.interp(x,size=self.size,mode=self.mode,align_corners=False)
         return x
         
 
@@ -235,12 +235,12 @@
 	
     def forward(self, x):
         
-        h = x.size()[2]
-        w = x.size()[3]
+        h = int(x.size()[2])
+        w = int(x.size()[3])
         
         b1 = self.branch1(x)
         # b1 = Interpolate(size=(h, w), mode="bilinear")(b1)
-        b1= interpolate(b1, size=(h, w), mode="bilinear", align_corners=True)
+        b1= interpolate(b1, size=(h, w), mode="bilinear", align_corners=False)
 	
         mid = self.mid(x)
 		
@@ -248,16 +248,18 @@
         x2 = self.down2(x1)
         x3 = self.down3(x2)
         # x3 = Interpolate(size=(h // 4, w // 4), mode="bilinear")(x3)
-        x3= interpolate(x3, size=(h // 4, w // 4), mode="bilinear", align_corners=True)	
+        x3= interpolate(x3, size=(h // 4, w // 4), mode="bilinear",
+                        align_corners=False)
         x2 = self.conv2(x2)
         x = x2 + x3
         # x = Interpolate(size=(h // 2, w // 2), mode="bilinear")(x)
-        x= interpolate(x, size=(h // 2, w // 2), mode="bilinear", align_corners=True)
+        x= interpolate(x, size=(h // 2, w // 2), mode="bilinear",
+                       align_corners=False)
        		
         x1 = self.conv1(x1)
         x = x + x1
         # x = Interpolate(size=(h, w), mode="bilinear")(x)
-        x= interpolate(x, size=(h, w), mode="bilinear", align_corners=True)
+        x= interpolate(x, size=(h, w), mode="bilinear", align_corners=False)
         		
         x = torch.mul(x, mid)
 
@@ -282,7 +284,8 @@
     def forward(self, input):
         
         output = self.apn(input)
-        out = interpolate(output, size=(512, 1024), mode="bilinear", align_corners=True)
+        out = interpolate(output, size=(512, 1024), mode="bilinear",
+                          align_corners=False)
         # out = self.upsample(output)
         # print(out.shape)
         return out
diff -x .git -x .gitignore -x '*.png' -r -N -r -u ./train/to_onnx.py esanet/src/models/external_code/LEDNet/train/to_onnx.py
--- ./train/to_onnx.py	1970-01-01 01:00:00.000000000 +0100
+++ esanet/src/models/external_code/LEDNet/train/to_onnx.py	2020-12-04 17:56:52.000000000 +0100
@@ -0,0 +1,27 @@
+import os
+import torch
+from lednet import Net
+
+N_CLASSES = 19
+H = 512
+W = 1024
+
+model = Net(N_CLASSES)
+model.eval()
+
+rgb = torch.rand(size=(1, 3, H, W), dtype=torch.float32)
+
+out_dir = '../../onnx_models'
+os.makedirs(out_dir, exist_ok=True)
+onnx_file_path = os.path.join(out_dir, 'lednet.onnx')
+
+torch.onnx.export(model,
+                  rgb,
+                  onnx_file_path,
+                  export_params=True,
+                  input_names=['rgb'],
+                  output_names=['output'],
+                  do_constant_folding=True,
+                  verbose=False,
+                  opset_version=10
+                  )