[FasterRCNN] fix type mismatch (#494)

examples/FasterRCNN/model.py

@@ -66,8 +66,8 @@ def rpn_losses(anchor_labels, anchor_boxes, label_logits, box_logits):
     with tf.device('/cpu:0'):
         valid_mask = tf.stop_gradient(tf.not_equal(anchor_labels, -1))
         pos_mask = tf.stop_gradient(tf.equal(anchor_labels, 1))
-        nr_valid = tf.stop_gradient(tf.count_nonzero(valid_mask), name='num_valid_anchor')
-        nr_pos = tf.count_nonzero(pos_mask, name='num_pos_anchor')
+        nr_valid = tf.stop_gradient(tf.count_nonzero(valid_mask, dtype=tf.int32), name='num_valid_anchor')
+        nr_pos = tf.count_nonzero(pos_mask, dtype=tf.int32, name='num_pos_anchor')
 
         valid_anchor_labels = tf.boolean_mask(anchor_labels, valid_mask)
     valid_label_logits = tf.boolean_mask(label_logits, valid_mask)
@@ -79,14 +79,16 @@ def rpn_losses(anchor_labels, anchor_boxes, label_logits, box_logits):
             for th in [0.5, 0.2, 0.1]:
                 valid_prediction = tf.cast(valid_label_prob > th, tf.int32)
                 nr_pos_prediction = tf.reduce_sum(valid_prediction, name='num_pos_prediction')
-                pos_prediction_corr = tf.count_nonzero(tf.logical_and(
-                    valid_label_prob > th,
-                    tf.equal(valid_prediction, valid_anchor_labels)))
+                pos_prediction_corr = tf.count_nonzero(
+                    tf.logical_and(
+                        valid_label_prob > th,
+                        tf.equal(valid_prediction, valid_anchor_labels)),
+                    dtype=tf.int32)
                 summaries.append(tf.truediv(
                     pos_prediction_corr,
                     nr_pos, name='recall_th{}'.format(th)))
-                precision = tf.truediv(pos_prediction_corr, nr_pos_prediction)
-                precision = tf.where(tf.equal(nr_pos_prediction, 0), 0, precision, name='precision_th{}'.format(th))
+                precision = tf.to_float(tf.truediv(pos_prediction_corr, nr_pos_prediction))
+                precision = tf.where(tf.equal(nr_pos_prediction, 0), 0.0, precision, name='precision_th{}'.format(th))
                 summaries.append(precision)
 
     label_loss = tf.nn.sigmoid_cross_entropy_with_logits(