add examples

app.py

@@ -26,6 +26,97 @@ MARKDOWN = """
 </div>
 """
 
+def init_clip_examples():
+    return gr.Examples(
+        [
+            [
+                "https://raw.githubusercontent.com/yu-rp/asserts/main/API_Prompting/clip_example_1_ori.jpg",
+                "https://raw.githubusercontent.com/yu-rp/asserts/main/API_Prompting/clip_example_1_masked.webp",
+                "What fruit is in the left part of the fridge?",
+                "-",
+                "fruit",
+                30,
+                3,
+                "BICUBIC",
+                0,
+                "On the left side of the fridge, there are strawberries and blueberries visible in containers.",
+                "In the left part of the fridge, there are strawberries in a clear plastic container."
+            ],
+            [
+                "https://raw.githubusercontent.com/yu-rp/asserts/main/API_Prompting/example_2_ori.jpg",
+                "https://raw.githubusercontent.com/yu-rp/asserts/main/API_Prompting/clip_example_2_masked.webp",
+                "On the right desk, what is to the left of the laptop?",
+                "-",
+                "desk",
+                5,
+                3,
+                "BICUBIC",
+                0,
+                "On the right desk, to the left of the laptop, there is a small pile of books stacked next to the laptop.",
+                "To the left of the laptop on the right desk, there is a desk lamp."
+            ],
+        ],
+        [
+            image_input, 
+            image_output, 
+            text_query, 
+            text_pre_answer, 
+            text_highlight_text, 
+            slider_enhance_coe,
+            slider_kernel_size,
+            radio_interpolate_method_name,
+            slider_mask_grayscale,
+            text_original_image_response,
+            text_API_image_response
+            ],
+        label = "Examples for CLIP_Based API"
+        )
+def init_llava_examples():
+    return gr.Examples(
+        [
+            [
+                "https://raw.githubusercontent.com/yu-rp/asserts/main/API_Prompting/llava_example_1_ori.jpg",
+                "https://raw.githubusercontent.com/yu-rp/asserts/main/API_Prompting/llava_example_1_masked.webp",
+                "Which direction is the red bird facing?",
+                " The red bird is facing away from the camera.</s>",
+                "red bird",
+                20,
+                3,
+                "BICUBIC",
+                100,
+                "The red bird in the image is facing to the right.",
+                "The red bird is facing towards the left in the image."
+            ],
+            [
+                "https://raw.githubusercontent.com/yu-rp/asserts/main/API_Prompting/example_2_ori.jpg",
+                "https://raw.githubusercontent.com/yu-rp/asserts/main/API_Prompting/llava_example_2_masked.webp",
+                "On the right desk, what is to the left of the laptop?",
+                " On the right desk, there is a lamp to the left of the laptop.</s>",
+                " On the right desk, there is a lamp to the left of the laptop.</s>",
+                30,
+                3,
+                "BICUBIC",
+                0,
+                "On the right desk, to the left of the laptop, there is a small pile of books stacked next to the laptop.",
+                "To the left of the laptop on the right desk, there is a green desk lamp."
+            ],
+        ],
+        [
+            image_input, 
+            image_output, 
+            text_query, 
+            text_pre_answer, 
+            text_highlight_text, 
+            slider_enhance_coe,
+            slider_kernel_size,
+            radio_interpolate_method_name,
+            slider_mask_grayscale,
+            text_original_image_response,
+            text_API_image_response
+            ],
+        label = "Examples for LLaVA_Based API"
+        )
+
 def get_base64_images(image):
     image = image.convert('RGB')
     buffer = BytesIO()
@@ -100,12 +191,14 @@ def change_api_method(api_method):
     if api_method == "CLIP_Based API":
         model_dict = init_clip()
         new_generate_llava_response_button = gr.Button("Pre-Answer", interactive=False)
+        # new_examples = init_clip_examples()
     elif api_method == "LLaVA_Based API":
         model_dict = init_llava()
         new_generate_llava_response_button = gr.Button("Pre-Answer", interactive=True)
+        # new_examples = init_llava_examples()
     else:
         raise NotImplementedError
-    return model_dict, {}, new_generate_llava_response_button, new_text_pre_answer, new_image_output
+    return model_dict, {}, new_generate_llava_response_button, new_text_pre_answer, new_image_output#, new_examples
 
 def clear_cache(cache_dict):
     return {}
@@ -193,7 +286,7 @@ text_pre_answer = gr.Textbox(
     interactive=False,
     type="text")
 text_highlight_text = gr.Textbox(
-    label = "Hint Text.",
+    label = "Hint Text",
     info = "The text based on which the mask will be generated. For LLaVA-Based API, it should be a substring of the pre-answer.",
     placeholder="Enter the hint text",
     lines=1,
@@ -286,24 +379,8 @@ with gr.Blocks() as demo:
             text_original_image_response.render()
             text_API_image_response.render()
     with gr.Accordion("Examples"):  
-        examples_images_responses = gr.Examples(
-            [
-
-            ],
-            [
-                image_input, 
-                image_output, 
-                text_query, 
-                text_pre_answer, 
-                text_highlight_text, 
-                slider_enhance_coe,
-                slider_kernel_size,
-                radio_interpolate_method_name,
-                slider_mask_grayscale,
-                text_original_image_response,
-                text_API_image_response
-                ],
-            )
+        clip_examples = init_clip_examples()
+        llava_examples = init_llava_examples()
 
     radio_api_method.change(
         fn=change_api_method,