Update distributionFocalLoss.m

Author: viakkala

+helper/distributionFocalLoss.m

@@ -1,7 +1,7 @@
-function boxes = distributionFocalLoss(boxes)
+function boxes = distributionFocalLoss(boxesInp)
 % Distribution Focal Loss Module
 
-sz = size(boxes);
+sz = size(boxesInp);
 c1=16; % pre-specified
 
 % compute batch. channel and anchors
@@ -14,43 +14,31 @@
 anchors = sz(2);
 
 % Reshape Operation
-boxes = permute(boxes,[2,1,3,4]); % 1 d qual like python
-boxes = reshape(boxes,anchors,c1,4,batch);
-boxes = permute(boxes,[2,1,3,4]);
+boxesInp = permute(boxesInp,[2,1,3,4]); 
+boxesReshaped = reshape(boxesInp,anchors,c1,4,batch);
+boxesMapped = permute(boxesReshaped,[2,1,3,4]);
 
 % Transpose Operation
-boxes = permute(boxes,[3,2,1,4]);
+boxesTrans = permute(boxesMapped,[3,2,1,4]);
+boxesTrans = extractdata(boxesTrans);
 
 % softmax along the channel dimension
-boxes = softmax(dlarray(boxes,'SSC'));  % produces a diff of 10^-3
-boxes = extractdata(boxes);
+boxesMax = softmax(dlarray(boxesTrans,'SSC')); 
+boxesMax = extractdata(boxesMax);
 
 % 1-d conv operation
 % Define weights
 weights = [0:c1-1];
-m = size(boxes,1);
-n = size(boxes,2);
+m = size(boxesMax,1);
+n = size(boxesMax,2);
 weights = reshape(repmat(weights,m*n,1),m,n,[]);
 % Conv operation
-boxes = boxes .* weights;
-boxes = sum(boxes,3);   % diff of 10^-2 because of above diff
+boxesConv = boxesMax .* weights;
+boxesTotal = sum(boxesConv,3); 
 
 % Reshape Operation
-boxes = permute(boxes,[2,1,3,4]);
-boxes = reshape(boxes,anchors,4,batch);
-boxes = permute(boxes,[2,1,3,4]);
+boxesTmp = permute(boxesTotal,[2,1,3,4]);
+boxesTmpReshaped = reshape(boxesTmp,anchors,4,batch);
+boxes = permute(boxesTmpReshaped,[2,1,3,4]);
 
 end
-
-
-
-
-
-
-
-
-
-
-
-
-