wip

app.py

@@ -52,8 +52,8 @@ OCR_WITH_REGION_EXAMPLES = [
     ["microsoft/Florence-2-large-ft", OCR_WITH_REGION_TASK_NAME, "https://media.roboflow.com/inference/license_plate_1.jpg"]
 ]
 
-# DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-DEVICE = "cuda"
+DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+# DEVICE = "cuda"
 MODELS, PROCESSORS = load_models(DEVICE)
 
 
@@ -74,10 +74,6 @@ def process(checkpoint_dropdown, task_dropdown, image_input):
         return response[task]
 
 
-image_output_component = None
-text_output_component = None
-
-
 with gr.Blocks() as demo:
     gr.Markdown(MARKDOWN)
     with gr.Row():
@@ -96,80 +92,19 @@ with gr.Blocks() as demo:
             submit_button_component = gr.Button(value='Submit', variant='primary')
 
         with gr.Column():
-            @gr.render(inputs=task_dropdown_component)
-            def show_output(text):
-                global image_output_component
-                global text_output_component
-                if text in [OBJECT_DETECTION_TASK_NAME, OCR_WITH_REGION_TASK_NAME]:
-                    image_output_component = gr.Image(type='pil', label='Image Output')
-                    submit_button_component.click(
-                        fn=process,
-                        inputs=[
-                            checkpoint_dropdown_component,
-                            task_dropdown_component,
-                            image_input_component
-                        ],
-                        outputs=image_output_component
-                    )
-                elif text in CAPTION_TASK_NAMES or text == OCR_TASK_NAME:
-                    text_output_component = gr.Textbox(label='Caption Output')
-                    submit_button_component.click(
-                        fn=process,
-                        inputs=[
-                            checkpoint_dropdown_component,
-                            task_dropdown_component,
-                            image_input_component
-                        ],
-                        outputs=text_output_component
-                    )
-
-    @gr.render(inputs=task_dropdown_component)
-    def show_examples(text):
-        global image_output_component
-        global text_output_component
-        if text == OBJECT_DETECTION_TASK_NAME:
-            gr.Examples(
-                fn=process,
-                examples=OBJECT_DETECTION_EXAMPLES,
-                inputs=[
-                    checkpoint_dropdown_component,
-                    task_dropdown_component,
-                    image_input_component
-                ],
-                outputs=image_output_component
-            )
-        elif text in CAPTION_TASK_NAMES:
-            gr.Examples(
-                fn=process,
-                examples=CAPTION_EXAMPLES,
-                inputs=[
-                    checkpoint_dropdown_component,
-                    task_dropdown_component,
-                    image_input_component
-                ],
-                outputs=text_output_component
-            )
-        elif text == OCR_TASK_NAME:
-            gr.Examples(
-                fn=process,
-                examples=OCR_EXAMPLES,
-                inputs=[
-                    checkpoint_dropdown_component,
-                    task_dropdown_component,
-                    image_input_component
-                ],
-                outputs=text_output_component
-            )
-        elif text == OCR_WITH_REGION_TASK_NAME:
-            gr.Examples(
-                fn=process,
-                examples=OCR_WITH_REGION_EXAMPLES,
-                inputs=[
-                    checkpoint_dropdown_component,
-                    task_dropdown_component,
-                    image_input_component
-                ],
-                outputs=image_output_component
-            )
+            image_output_component = gr.Image(type='pil', label='Image Output')
+            text_output_component = gr.Textbox(label='Caption Output', visible=False)
+
+    def on_dropdown_input(text):
+        if text in CAPTION_TASK_NAMES + [OCR_TASK_NAME]:
+            return [gr.Image(visible=False), gr.Textbox(visible=True)]
+        else:
+            return [gr.Image(visible=True), gr.Textbox(visible=False)]
+
+    task_dropdown_component.input(
+        on_dropdown_input,
+        inputs=[task_dropdown_component],
+        outputs=[image_output_component, text_output_component])
+
 
 demo.launch(debug=False, show_error=True, max_threads=1)