update

app.py

@@ -4,7 +4,7 @@ from scorer import DSGPromptProcessor
 import matplotlib.pyplot as plt
 import networkx as nx
 from PIL import Image
-
+import io
 
 def draw_colored_graph(dependencies, questions, answers):
     # Create a directed graph
@@ -52,7 +52,7 @@ def process_image(image, prompt):
     )
     reward = processor.get_reward(questions, dependencies, [image])
     reward = reward[0]
-    answers = {i: v > 0.5 for i, v in enumerate(reward)}
+    answers = {str(i): v > 0.5 for i, v in enumerate(reward)}
     graph_img = draw_colored_graph(dependencies, questions, answers)
     return reward, f"""
 Question: {questions}.
@@ -63,7 +63,8 @@ Reward per question: {reward}"""
 description = """
 <p align="center">T2I Adherence Scorer based on Davidsonian Scene Graph</p>
 <p><center>
-<a >Original paper: Davidsonian Scene Graph: Improving Reliability in Fine-grained Evaluation for Text-to-Image Generation (ICLR 2024)</a>
+<a href="https://arxiv.org/abs/2310.18235A" target="_blank">[Original Paper]</a>
+<a href="https://github.com/toilaluan" target="_blank">[My Github]</a>
 <a href="https://huggingface.co/toilaluan/Florence-2-base-Yes-No-VQA" target="_blank">[Binary VQA Model - Query Answering]</a>
 <a href="https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1" target="_blank">[Mixtral 7x8 - Query Generating]</a>
 </center></p>
@@ -72,8 +73,8 @@ description = """
 # Define the Gradio interface
 interface = gr.Interface(
     fn=process_image, 
-    inputs=[gr.Image(type="pil"), gr.Textbox(label="Enter your prompt")], 
-    outputs=[gr.Image(type="pil"), gr.Textbox(label="Output text")],
+    inputs=[gr.Image(type="pil", label="Input Image"), gr.Textbox(label="Enter your prompt")], 
+    outputs=[gr.Image(type="pil", label="Graph Score Image"), gr.Textbox(label="Analyzed Result")],
     theme=gr.themes.Soft(),
     title="",
     description=description