Update app.py

app.py

@@ -40,7 +40,7 @@ if not precomputed_results:
     device = "cuda" if torch.cuda.is_available() else "cpu"
     model = model.to(device)
 
-def change_language(lang, randomize_imgs):
+def change_language(lang):
     # compute text embeddings
     labels = babel_imagenet[lang][1]
     class_order = list(range(len(labels)))
@@ -78,7 +78,7 @@ def select(idx, lang, choice, correct, model_choice, player_score, clip_score, c
 
     return correct_text, player_score_text, clip_score_text, player_score, clip_score
 
-def prepare(raw_idx, lang, text_embeddings, class_order, randomize_images):
+def prepare(raw_idx, lang, text_embeddings, class_order):
     #  prepared next question, loads image, and computes choices
 
     raw_idx = (raw_idx+1) % len(babel_imagenet[lang][0])
@@ -94,8 +94,7 @@ def prepare(raw_idx, lang, text_embeddings, class_order, randomize_images):
         class_idx = lang_class_idxs[idx]
 
     img_idx = 0
-    if randomize_images:
-        img_idx = np.random.choice(min(len(babelnet_images[class_idx]), max_image_choices))
+    img_idx = np.random.choice(min(len(babelnet_images[class_idx]), max_image_choices))
     img_url = babelnet_images[class_idx][img_idx]["url"]
     class_labels = babel_imagenet[lang][1] if lang != "EN" else openai_en_classes
 
@@ -107,7 +106,7 @@ def prepare(raw_idx, lang, text_embeddings, class_order, randomize_images):
                 image_features /= image_features.norm(dim=-1, keepdim=True)
         except:
             gr.Warning("There is a problem with the next class. Skipping it.")
-            return prepare(raw_idx, lang, text_embeddings, class_order, randomize_images)
+            return prepare(raw_idx, lang, text_embeddings, class_order)
 
         similarity = (text_embeddings @ image_features.cpu().numpy().T).squeeze()
         choices = np.argsort(similarity)[-4:].tolist()
@@ -135,7 +134,7 @@ def prepare(raw_idx, lang, text_embeddings, class_order, randomize_images):
     return next_radio, next_image, raw_idx, correct_choice, model_choice, choice_values
 
 
-def reroll(raw_idx, lang, text_embeddings, class_order, randomize_images):
+def reroll(raw_idx, lang, text_embeddings, class_order):
     #  prepared next question, loads image, and computes choices
 
     idx = class_order[raw_idx]
@@ -143,8 +142,7 @@ def reroll(raw_idx, lang, text_embeddings, class_order, randomize_images):
     class_idx = lang_class_idxs[idx]
 
     img_idx = 0
-    if randomize_images:
-        img_idx = np.random.choice(min(len(babelnet_images[class_idx]), max_image_choices))
+    img_idx = np.random.choice(min(len(babelnet_images[class_idx]), max_image_choices))
     img_url = babelnet_images[class_idx][img_idx]["url"]
     class_labels = babel_imagenet[lang][1] if lang != "EN" else openai_en_classes
 
@@ -156,7 +154,7 @@ def reroll(raw_idx, lang, text_embeddings, class_order, randomize_images):
                 image_features /= image_features.norm(dim=-1, keepdim=True)
         except:
             gr.Warning("There is a problem with the next class. Skipping it.")
-            return prepare(raw_idx, lang, text_embeddings, class_order, randomize_images)
+            return prepare(raw_idx, lang, text_embeddings, class_order)
 
         similarity = (text_embeddings @ image_features.cpu().numpy().T).squeeze()
         choices = np.argsort(similarity)[-4:].tolist()
@@ -199,16 +197,15 @@ with (gr.Blocks(title="Babel-ImageNet Quiz") as demo):
 
     # Title Area
     gr.Markdown("""
-# Are you smarter🤓 than CLIP🤖?
+# ¿Eres más lista 🤓 que CLIP 🤖?
 
-<small>by Gregor Geigle, WüNLP & Computer Vision Lab, University of Würzburg</small>
+<small>por Gregor Geigle, WüNLP & Computer Vision Lab, University of Würzburg y Both.rocks!</small>
 
         """)
     # language select dropdown
     with gr.Row():
         language_select = gr.Dropdown(choices=main_language_values, value="EN", interactive=True, label="Select your language:")
-        randomize_images = gr.Checkbox(label="Randomize images (if unchecked, will always show the same image). Other images might be less relevant.", value=True)
-        start_btn = gr.Button(value="Start", variant="primary")
+        start_btn = gr.Button(value="¡Adelante!", variant="primary")
 
     # quiz area
     with gr.Row():
@@ -217,9 +214,9 @@ with (gr.Blocks(title="Babel-ImageNet Quiz") as demo):
                              width=IMG_WIDTH, height=IMG_WIDTH)
 
         with gr.Column(scale=1):
-            options = gr.Radio(choices=["Click", "start", "to", "begin"], interactive=False, label="Please click start to begin.")
+            options = gr.Radio(choices=["", "", "", ""], interactive=False, label="Pulsa ¡Adelante! para empezar.")
             # with gr.Row():
-            correct_text = gr.Text("Please click start to begin.")
+            correct_text = gr.Text("Pulsa ¡Adelante! para empezar.")
             player_score_text = gr.Text(f"Player score: 0")
             clip_score_text = gr.Text(f"mSigLIP score: 0")
             reroll_btn = gr.Button(value="Reroll the image (for bad images or errors)")
@@ -231,18 +228,18 @@ with (gr.Blocks(title="Babel-ImageNet Quiz") as demo):
                    inputs=[class_idx, language_select, options, correct_choice, model_choice, player_score, clip_score, choices],
                    outputs=[correct_text, player_score_text, clip_score_text, player_score, clip_score]
                    ).then(fn=prepare,
-                          inputs=[class_idx, language_select, text_embeddings, class_order, randomize_images],
+                          inputs=[class_idx, language_select, text_embeddings, class_order],
                           outputs=[options, image, class_idx, correct_choice, model_choice, choices])
 
     start_btn.click(fn=change_language,
-                          inputs=[language_select, randomize_images],
+                          inputs=[language_select],
                           outputs=[text_embeddings, class_idx, class_order, correct_text, player_score_text, clip_score_text, player_score, clip_score]
                           ).then(fn=prepare,
-                                 inputs=[class_idx, language_select, text_embeddings, class_order, randomize_images],
+                                 inputs=[class_idx, language_select, text_embeddings, class_order],
                                  outputs=[options, image, class_idx, correct_choice, model_choice, choices])
 
     reroll_btn.click(fn=reroll,
-                     inputs=[class_idx, language_select, text_embeddings, class_order, randomize_images],
+                     inputs=[class_idx, language_select, text_embeddings, class_order],
                      outputs=[options, image, class_idx, correct_choice, model_choice, choices])