Ricommenti per la dic_history

README.md

@@ -84,7 +84,7 @@ Specify the mode to start the framework. The list of MODE value is given below:
         <li>train_cDCGAN: train the conditional Generative Adversarial Network </li>
         <li>evaluate_G1: calculate FID and IS scores of G<sub>1</sub> </li>
         <li>evaluate_GAN: calculate FID and IS scores of all framework</li>
-        <li>tsne_GAN: calculate t-sne of all framework </li>
+        <li>tsne_GAN: calculate t-sne of all framework</li>
         <li>inference_G1: inference on test set using G<sub>1</sub> </li>
         <li>inference_GAN: inference on test set using all framework</li>
 		<li>plot_history_G1:  plot history file of G<sub>1</sub> training</li>

data/README.md

@@ -37,6 +37,7 @@ The content of each folder is described below:
 	<ul>
 		<li>train/valid/test sets in <i>.tfrecord</i> format. These are the set to use during the training and evaluation phase</li>
 		<li>sets_configs.pkl that describe the carateristics about the configuration <i>(radius_key=2, flip=True, etc..)</i></li>
+		<li>dic_history.pkl dictonary in which we have for each sets the pair formed and the related positin in tfrecord file</li>
 	</ul>
 	These files are created by the <a href="./Dataset_configuration_generator.py"> ./Dataset_configuration_generator.py </a> script described in section 2.</li>
 </ul>

data/Syntetich_complete/README.md

@@ -27,6 +27,7 @@ In each of configuration directory there are:
 <ul>
 <li> the train, valid e test sets files in .tfrecord format</li>
 <li> sets_config file in pikle (.pkl) format that containg the infos on configuration. </li>
+<li> dic_history.pkl dictonary in which we have for each sets the pairs formed and the related position in the tfrecord file</li>
 </ul>
 For the Syntetich complete, we created only one configuration called: negative_no_flip_camp_5_keypoints_2_mask_1.
 The configuration and related files are created by the 

src/main.py

@@ -21,7 +21,11 @@
         pg2.evaluate_GAN(analysis_set="test_set", name_dataset=config.name_tfrecord_test,
                          dataset_len=config.dataset_test_len)
     elif config.MODE == 'tsne_GAN':
-        pg2.tsne(key_image_interested="test_20")  # id immagine ottenuto dalla inference
+        # The dic_history_key_pair refers to a specific pair.
+        # In this case, 'test_20', signify that we want compare the real and generated features of
+        # 20th pair in test set. You can write other pairs. In particular, the value that you can use are defined in
+        # dic_history.pkl file of your dataset configuration as keys of dictionary
+        pg2.tsne(dic_history_key_pair="test_20")
     elif config.MODE == 'inference_G1':
         pg2.inference_on_test_set_G1()
     elif config.MODE == 'inference_GAN':

src/pg2.py

@@ -968,7 +968,7 @@ def plot_history_GAN(self):
 
         plt.show()
 
-    def tsne(self, key_image_interested="test_20"):
+    def tsne(self, dic_history_key_pair="test_20"):
         self.config.load_train_path_G1()
         self.config.load_train_path_GAN()
         tsne_path = os.path.join(self.config.OUTPUTS_DIR, "evaluation", "tsne")
@@ -998,4 +998,4 @@ def tsne(self, key_image_interested="test_20"):
         utils.vgg16_pca_tsne_features.start(list_sets, list_perplexity,
                                             self.G1, self.G2, self.dataset_module,
                                             dir_to_save=tsne_path, save_fig_plot=True,
-                                            key_image_interested=key_image_interested)
+                                            key_image_interested=dic_history_key_pair)