Update app.py

app.py

@@ -1,8 +1,8 @@
-import os
 import requests 
 from PIL import Image, UnidentifiedImageError
 from transformers import AutoTokenizer, AutoImageProcessor, VisionEncoderDecoderModel
 import gradio as gr
+import os
 
 # Load the model, tokenizer, and image processor with error handling
 def load_model_and_components(model_name):
@@ -27,32 +27,16 @@ def generate_caption(image):
     except Exception:
         return "Please upload a valid image."
 
-# Predefined images for selection
-predefined_images_urls = [
-    "http://images.cocodataset.org/val2014/COCO_val2014_000000458153.jpg",
-    "http://images.cocodataset.org/val2014/COCO_val2014_000000000074.jpg"
-]
-
 # Predefined images for selection
 image_folder = "images"
 predefined_images_paths = [
     os.path.join(image_folder, fname) for fname in os.listdir(image_folder) if fname.lower().endswith(('.png', '.jpg', '.jpeg', '.bmp', '.gif'))
 ]
 
-# Preload PIL.Image objects for predefined images with error handling
-predefined_images = []
-for path in predefined_images_paths:
-    try:
-        predefined_images.append(Image.open(path))
-    except Exception as e:
-        print(f"Error loading predefined image from {path}: {e}")
-
 # Gradio app
-def app(image=None, predefined_image=None):
+def app(image=None):
     try:
-        if predefined_image is not None:
-            image = predefined_image
-        elif image is None:
+        if image is None:
             return "Please upload a valid image."
         return generate_caption(image)
     except Exception:
@@ -60,11 +44,6 @@ def app(image=None, predefined_image=None):
 
 # Define UI
 with gr.Blocks() as interface:
-    gr.Markdown("""
-        # Welcome to the LAICSI-IFES space for Vision Encoder-Decoder (VED) demonstration
-
-        ### Be patient with the Swin-GPorTuguese-2 as it is heavier than the Swin-DistilBERTimbau.
-    """)
     with gr.Row():
         with gr.Column():
             model_selector = gr.Dropdown(choices=["laicsiifes/swin-distilbertimbau", "laicsiifes/swin-gportuguese-2"], 
@@ -73,7 +52,7 @@ with gr.Blocks() as interface:
             loading_message = gr.Textbox(label="Status Message")
             image_display = gr.Image(type="pil", label="Image Preview", interactive=False)
             upload_button = gr.File(label="Upload an Image", file_types=["image"], type="filepath")
-            predefined_images_display = gr.Gallery(predefined_images_urls, label="Choose a Predefined Image")
+            examples = gr.Examples(predefined_images_paths, inputs=[upload_button], label="Examples")
 
         with gr.Column():
             output_text = gr.Textbox(label="Generated Caption")
@@ -88,15 +67,6 @@ with gr.Blocks() as interface:
         except Exception:
             return None, "Please upload a valid image."
 
-    def handle_predefined_image(evt: gr.SelectData, _):
-        try:
-            if not evt:
-                return None, "Please upload a valid image."
-            pil_image = Image.open(requests.get(evt.value['image']['url'], stream=True).raw)
-            return pil_image, generate_caption(pil_image)
-        except Exception:
-            return None, "Please upload a valid image."
-
     def switch_model(selected_model):
         gr.Info("Loading model... Please wait.")
         return "Loading model... Please wait.", None, None, None
@@ -107,8 +77,7 @@ with gr.Blocks() as interface:
         return "Model loaded successfully.", None, None, None
 
     model_selector.change(fn=switch_model, inputs=model_selector, outputs=[loading_message, upload_button, image_display, output_text])
-    model_selector.change(fn=load_new_model, inputs=model_selector, outputs=[loading_message, upload_button, image_display, output_text])
+    model_selector.change(fn=load_new_model, inputs=model_selector, outputs=[loading_message, image_display, output_text])
     upload_button.change(fn=handle_uploaded_image, inputs=upload_button, outputs=[image_display, output_text])
-    predefined_images_display.select(fn=handle_predefined_image, inputs=predefined_images_display, outputs=[image_display, output_text])
 
-    interface.launch()
+    interface.launch(share=False)