bn in answer inference

answer_classifier_cached_features/inference.py

@@ -95,9 +95,10 @@ class AnswerInference():
                     reuse_vars = reuse_vars)
 
                 self.per_region_answer_scores[j] = tf.nn.relu(
-                    self.batch_norm(
+                    layers.batch_norm(
                         self.per_region_answer_scores[j],
-                        self.is_training))                
+                        tf.constant(self.is_training),
+                        reuse_vars = reuse_vars))                
 
                 self.per_region_answer_scores[j] = layers.conv2d(
                     self.per_region_answer_scores[j],

answer_classifier_cached_features/train.py

@@ -551,6 +551,8 @@ def create_vqa_batch_generator():
         constants.vqa_train_subset_qids,
         constants.vocab_json,
         constants.vqa_answer_vocab_json,
+        constants.object_labels_json,
+        constants.attribute_labels_json,
         constants.image_size,
         constants.num_region_proposals,
         constants.num_negative_answers,
@@ -850,6 +852,8 @@ if __name__=='__main__':
         constants.num_object_labels,
         constants.num_attribute_labels,
         constants.answer_obj_atr_loss_wt,
+        constants.answer_ans_loss_wt,
+        constants.answer_mil_loss_wt,
         resnet_feat_dim=constants.resnet_feat_dim,
         training=True)
 

constants_crunchy.py

@@ -5,7 +5,7 @@ def mkdir_if_not_exists(dir_name):
     if not os.path.exists(dir_name):
         os.mkdir(dir_name)
         
-experiment_name = 'QA_classifier_joint_pretrain_wordvec_xform_select' #'QA_joint_pretrain_genome_split'
+experiment_name = 'class_through_vqa_mil_ans' #'QA_joint_pretrain_genome_split'
 
 # Global output directory (all subexperiments will be saved here)
 global_output_dir = '/home/tanmay/Code/GenVQA/Exp_Results/VQA'

data/vqa_cached_features.py

@@ -107,9 +107,9 @@ class data():
         batch['negative_answers_unencoded'] = negative_answers_unencoded
 
         batch['positive_nouns'] = batch['question_nouns'] \
-                                  + batch['positive_answer_nouns'])
+                                  + batch['positive_answer_nouns']
         batch['positive_adjectives'] = batch['question_adjectives'] \
-                                       + batch['positive_answer_adjectives'])
+                                       + batch['positive_answer_adjectives']
 
         _, batch['positive_nouns_vec_enc'] = self.noun_to_obj_id(
             batch['positive_nouns'],
@@ -155,8 +155,11 @@ class data():
 
         for i, id in enumerate(list_of_noun_ids):
             if id in self.inv_vocab:
-                obj_ids[i] = int(self.obj_labels[self.inv_vocab[id]])
-                vec_enc[obj_ids[i]] = 1.0
+                if self.inv_vocab[id] in self.obj_labels:
+                    obj_ids[i] = int(self.obj_labels[self.inv_vocab[id]])
+                    vec_enc[0,obj_ids[i]] = 1.0
+                else:
+                    obj_ids[i] = -1
             else:
                 obj_ids[i] = -1
                 
@@ -171,8 +174,9 @@ class data():
 
         for i, id in enumerate(list_of_adj_ids):
             if id in self.inv_vocab:
-                atr_ids[i] = int(self.atr_labels[self.inv_vocab[id]])
-                vec_enc[atr_ids[i]] = 1.0
+                if self.inv_vocab[id] in self.atr_labels:
+                    atr_ids[i] = int(self.atr_labels[self.inv_vocab[id]])
+                    vec_enc[0,atr_ids[i]] = 1.0
             else:
                 atr_ids[i] = -1