clean code

Author: yi.he

alignshift/operators/alignshiftconv.py

@@ -33,8 +33,8 @@ def __init__(self, in_channels, out_channels, kernel_size, stride=1,
         self.ref_thickness = ref_thickness
         self.shift_padding_zero = shift_padding_zero
 
-    def alignshift(self, input, fold, thickness, padding_zero=True):
-        alph = 1 - (self.ref_thickness / thickness).view(-1, 1 ,1, 1, 1).clamp_(0., 1.)
+    def alignshift(self, input, fold, ref_thickness, thickness, padding_zero=True):
+        alph = 1 - (ref_thickness / thickness).view(-1, 1 ,1, 1, 1).clamp_(0., 1.)
         out = torch.zeros_like(input)
         out[:, :fold, :-1] = input[:,  :fold, :-1] * alph +  input[:,  :fold, 1:] * (1-alph)
         out[:, fold: 2 * fold, 1:] = input[:,  fold: 2 * fold, 1:] * alph + \
@@ -49,7 +49,7 @@ def align_shift(self, x, fold, ref_thickness, thickness, padding_zero, inplace):
         if inplace:
             x = inplace_alignshift(x, fold, ref_thickness, thickness, padding_zero)
         else:
-            x = self.alignshift(x, fold, thickness, padding_zero)
+            x = self.alignshift(x, fold, ref_thickness, thickness, padding_zero)
         return x
 
     def forward(self, input, thickness=None):
@@ -67,7 +67,7 @@ def extra_repr(self):
 
 class InplaceAlignShift(torch.autograd.Function):
     @staticmethod
-    def forward(ctx, input, fold, align_spacing, thickness, padding_zero=True):
+    def forward(ctx, input, fold, ref_thickness, thickness, padding_zero=True):
         '''
         @params: 
             input: BxCxDxHxW
@@ -79,7 +79,7 @@ def forward(ctx, input, fold, align_spacing, thickness, padding_zero=True):
         n, c, t, h, w = input.size()
         ctx.fold_ = fold
         ctx.padding_zero = padding_zero
-        alph = 1 - (align_spacing / thickness).view(-1, 1 ,1, 1, 1).clamp_(0., 1.) ##把小于align_spacing的当作align_spacing，不做插值处理 
+        alph = 1 - (ref_thickness / thickness).view(-1, 1 ,1, 1, 1).clamp_(0., 1.) ##把小于ref_thickness的当作ref_thickness，不做插值处理 
         ctx.alph_ = alph
         input.data[:, :fold, :-1] = input.data[:, :fold, :-1] * alph + input.data[:, :fold, 1:] * (1-alph)
         input.data[:, fold:2*fold, 1:] = input.data[:, fold:2*fold, 1:] * alph + \

deeplesion/eval.py

@@ -173,7 +173,7 @@ def mask_matrics(output, iou_thresh=0.5):
                 # _, cnts = cv2.findContours(l_seg, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
                 pred_mask_index.append(indx)
                 prop = regionprops(d_seg)[0]
-                diameter = np.sqrt(hprop.major_axis_lengt**2 + prop.minor_axis_length**2)/2
+                diameter = np.sqrt(prop.major_axis_lengt**2 + prop.minor_axis_length**2)/2
                 recists = output['recists'][i].reshape((4, 2))
                 l = np.linalg.norm(recists[0] - recists[1])
                 s = np.linalg.norm(recists[2] - recists[3])