FixClustGUIAxes

Functions/Call Classification/CreateClusteringData.m

@@ -26,6 +26,8 @@
 freqRange = [];
 xFreq = [];
 xTime = [];
+FreqScale = [];
+TimeScale = [];
 stats.Power = [];
 
 % Select the files
@@ -94,7 +96,7 @@
     waitbar(i/height(Calls),h, sprintf('Loading File %u of %u', Calls.audiodata_index(i), length(fileName)));
 
     % Change the audio file if needed
-    if Calls.audiodata_index(i) ~= currentAudioFile;
+    if Calls.audiodata_index(i) ~= currentAudioFile
         audioReader.audiodata = audiodata{Calls.audiodata_index(i)};
         currentAudioFile = Calls.audiodata_index(i);
         perFileCallID = 0;
@@ -137,13 +139,16 @@
         {perFileCallID} % Call ID in file
         {stats.Power}
         {box(4)}
+        {FreqScale}
+        {TimeScale}
+        {Calls.CallID(i)}
         ]'];
 
     clustAssign = [clustAssign; Calls.Type(i)];
 end
 
 
-ClusteringData = cell2table(ClusteringData, 'VariableNames', {'Spectrogram', 'MinFreq', 'Duration', 'xFreq', 'xTime', 'Filename', 'callID', 'Power', 'Bandwidth'});
+ClusteringData = cell2table(ClusteringData(:,1:12), 'VariableNames', {'Spectrogram', 'MinFreq', 'Duration', 'xFreq', 'xTime', 'Filename', 'callID', 'Power', 'Bandwidth','FreqScale','TimeScale','UserID'});
 
 close(h)
 

Functions/Call Classification/UpdateCluster.m

@@ -29,8 +29,13 @@ function UpdateCluster(ClusteringData, clustAssign, clusterName, rejected)
     % Find the index of the clustering data that belongs to the file
     cluster_idx = find(file_idx == i);
 
-    % Find the index of the calls in the file that correspond the the clustering data
-    call_idx = [ClusteringData{cluster_idx, 7}];
+    if ismember('UserID',obj.ClusteringData.Properties.VariableNames)
+        % Find the index of the calls in the file that correspond the the clustering data
+        [~,call_idx] = ismember(ClusteringData{cluster_idx, 'UserID'},Calls.CallID);
+    else
+        warning('This will not assign clusters correctly if rejected Calls were not removed from your detections file.')
+        [~,call_idx] = ismember(ClusteringData{cluster_idx, 'callID'},Calls.CallID);
+    end
 
     % Update call type with cluster names
     Calls.Type(call_idx) = clustAssign(cluster_idx);

Functions/Call Classification/clusteringGUI.m

@@ -33,8 +33,14 @@
             obj.ClusteringData = ClusteringData;
             obj.rejected = zeros(1,length(obj.clustAssign));
 
-            obj.minfreq = prctile(ClusteringData.MinFreq, 5);
-            obj.maxfreq = prctile(ClusteringData.MinFreq + ClusteringData.Bandwidth, 95);
+%             obj.minfreq = prctile(ClusteringData.MinFreq, 5);
+%             obj.maxfreq = prctile(ClusteringData.MinFreq + ClusteringData.Bandwidth, 95);
+            %This is going to be scaled from 0 to the Nyquist, not the
+            %boxed call
+            obj.minfreq = 0;
+            %Assumes all calls in ClusteringData were recorded at the same
+            %SR
+            obj.maxfreq = size(ClusteringData.Spectrogram{1},1)*ClusteringData.FreqScale(1);
             obj.ColorData = jet(256); % Color by mean frequency
             % obj.ColorData = HSLuv_to_RGB(256, 'H',  [270 0], 'S', 100, 'L', 75, 'type', 'HSL'); % Make a color map for each category
             obj.ColorData = reshape(obj.ColorData,size(obj.ColorData,1),1,size(obj.ColorData,2));
@@ -220,8 +226,8 @@ function render_GUI(obj)
             % Resize the image while maintaining the aspect ratio by
             % padding with zeros
             im_size = size(ClusteringData.Spectrogram{clustIndex(callID)}) ;
-            new_size = floor(im_size .* min(obj.thumbnail_size ./ im_size));
-            im = double(imresize(ClusteringData.Spectrogram{clustIndex(callID)}, new_size));
+            new_size = obj.thumbnail_size;
+            im = double(imresize(ClusteringData.Spectrogram{clustIndex(callID)}, obj.thumbnail_size));
             pad = (obj.thumbnail_size - size(im)) / 2;
             im = padarray(im, floor(pad), 'pre');
             im = padarray(im, ceil(pad), 'post');
@@ -231,9 +237,12 @@ function render_GUI(obj)
             rel_x = [rel_size(2), 1-rel_size(2)];
             rel_y = [rel_size(1), 1-rel_size(1)];
 
-            % Apply color to the greyscale images
-            freqRange = [ClusteringData.MinFreq(clustIndex(callID)),...
-                ClusteringData.MinFreq(clustIndex(callID)) + ClusteringData.Bandwidth(clustIndex(callID))];
+%             % Apply color to the greyscale images
+                %this type of color scaling isn't applicable as is because
+                %all call images have the same frequency span
+%             freqRange = [ClusteringData.MinFreq(clustIndex(callID)),...
+%                 ClusteringData.MinFreq(clustIndex(callID)) + ClusteringData.Bandwidth(clustIndex(callID))];
+            freqRange = [obj.minfreq, obj.maxfreq];
             % Account for any padding on the y axis
             freqRange = freqRange + range(freqRange) .* rel_y(1) .* [-1, 1];
 
@@ -249,14 +258,16 @@ function render_GUI(obj)
             x_lim = xlim(axis_handles);
             x_span = x_lim(2) - x_lim(1);
             xtick_positions = linspace(x_span*rel_x(1)+x_lim(1), x_span*rel_x(2)+x_lim(1),4);
-            x_ticks = linspace(0,obj.ClusteringData.Duration(i),4);
+            %x_ticks = linspace(0,obj.ClusteringData.Duration(i),4);
+            x_ticks = linspace(0,size(obj.ClusteringData.Spectrogram{i},2)*obj.ClusteringData.TimeScale(i),4);
             x_ticks = arrayfun(@(x) sprintf('%.3f',x),x_ticks(2:end),'UniformOutput',false);
 
             y_lim = ylim(axis_handles);
             y_span = y_lim(2) - y_lim(1);
-            ytick_positions = linspace(y_span*rel_y(1)+y_lim(1), y_span*rel_y(2)+y_lim(1),3);            
+            ytick_positions = linspace(y_span*rel_y(1)+y_lim(1), y_span*rel_y(2)+y_lim(1),4);            
 
-            y_ticks = linspace(obj.ClusteringData.MinFreq(i),obj.ClusteringData.MinFreq(i)+obj.ClusteringData.Bandwidth(i),3);
+            %y_ticks = linspace(obj.ClusteringData.MinFreq(i),obj.ClusteringData.MinFreq(i)+obj.ClusteringData.Bandwidth(i),3);
+            y_ticks = linspace(obj.minfreq,obj.maxfreq,4);
             y_ticks = arrayfun(@(x) sprintf('%.1f',x),y_ticks(1:end),'UniformOutput',false);
             y_ticks = flip(y_ticks);
 

Functions/CreateFocusSpectrogram.m

@@ -38,7 +38,8 @@
     s  = handles.data.page_spect.s(:,handles.data.page_spect.t > call.Box(1) & handles.data.page_spect.t < sum(call.Box([1,3])));
     ti = handles.data.page_spect.t(  handles.data.page_spect.t > call.Box(1) & handles.data.page_spect.t < sum(call.Box([1,3])));
     fr = handles.data.page_spect.f;
-
+    p = (1/(rate*(hamming(nfft)'*hamming(nfft))))*abs(s).^2;
+    p(2:end-1,:) = p(2:end-1,:).*2;
 end
 
 %% Get the part of the spectrogram within the box

Functions/Network Training/TrainSqueakDetector.m

@@ -49,7 +49,7 @@
     [detector,info] = trainYOLOv2ObjectDetector(TrainingTables,lgraph,options);
 elseif nargin == 2
     [detector,info] = trainYOLOv2ObjectDetector(TrainingTables,layers,options);
-else
+else 
     error('This should not happen');   
 end
 end

Functions/loadCallfile.m

@@ -13,6 +13,13 @@
     error('This doesn''t appear to be a detection file!')
 end
 
+if ~any(strcmp('CallID', Calls.Properties.VariableNames)) || length(unique(Calls.CallID)) ~= height(Calls)
+    if length(unique(Calls.CallID)) ~= height(Calls)
+        warning('CallID not unique - replacing with 1:height(Calls)')
+    end
+    Calls.CallID = categorical(1:height(Calls))';
+end
+
 if isfield(data, 'audiodata')
     audiodata = data.audiodata;
 end