diff --git a/fixtracks/widgets/classifier.py b/fixtracks/widgets/classifier.py
index e610a29..8ed57fc 100644
--- a/fixtracks/widgets/classifier.py
+++ b/fixtracks/widgets/classifier.py
@@ -135,6 +135,16 @@ class ConsistencyWorker(QRunnable):
             del detections[lowest_dist]
             return detections
 
+        def find_last_userlabeled(startframe):
+            t1index = np.where((self.frames < startframe) & (self.userlabeled) & (self.tracks == 1))[0][-1]
+            t2index = np.where((self.frames < startframe) & (self.userlabeled) & (self.tracks == 2))[0][-1]
+            d1 = Detection(t1index, self.frames[t1index], self.tracks[t1index], self.positions[t1index],
+                           self.orientations[t1index], self.lengths[t1index], self.userlabeled[t1index])
+            d2 = Detection(t1index, self.frames[t2index], self.tracks[t2index], self.positions[t2index],
+                           self.orientations[t2index], self.lengths[t2index], self.userlabeled[t2index])
+            last_detections[1] = d1
+            last_detections[2] = d2
+
         unique_frames = np.unique(self.frames)
         steps = int((len(unique_frames) - self._startframe) // 100)
         errors = 0
@@ -142,6 +152,7 @@ class ConsistencyWorker(QRunnable):
         progress = 0
         self._stoprequest = False
         last_detections = {1: None, 2: None, -1: None}
+        find_last_userlabeled(self._startframe)
 
         for f in unique_frames[unique_frames >= self._startframe]:
             if self._stoprequest:
@@ -188,7 +199,7 @@ class ConsistencyWorker(QRunnable):
                 continue
 
             if error and self._stoponerror:
-                self.signals.message.emit("Tracking stopped at frame %i.", f)
+                self.signals.message.emit(f"Tracking stopped at frame {f}.")
                 break
             elif error:
                 continue
@@ -559,8 +570,11 @@ class ConsistencyClassifier(QWidget):
                 self._messagebox.append("Error preparing data! Make sure that the first user-labeled frames contain both tracks!")
                 self.setEnabled(False)
                 return
-            max_startframe = np.min([t1_userlabeled[-1], t2_userlabeled[-1]])
-            min_startframe = np.max([t1_userlabeled[0], t2_userlabeled[0]])
+            max_startframe = np.min([t1_userlabeled[-1], t2_userlabeled[-1]]) -1
+            first_guess = np.max([t1_userlabeled[0], t2_userlabeled[0]])
+            while first_guess not in t1_userlabeled or first_guess not in t2_userlabeled:
+                first_guess += 1
+            min_startframe = first_guess + 1
             self._maxframes = np.max(self._frames)
             self._maxframeslabel.setText(str(self._maxframes))
             self._startframe_spinner.setMinimum(min_startframe)