Update app.py

app.py

@@ -6,8 +6,17 @@ from PIL import Image
 from transformers import pipeline
 
 
+if "model" not in st.session_state:
+    st.session_state.model = pipeline("image-to-text", model="Salesforce/blip-image-captioning-large")
+
+if "result" not in st.session_state:
+    st.session_state.result = ""
+    
+if "essay" not in st.session_state:
+    st.session_state.essay = ""
+        
 # 初始化图片描述生成模型
-image_to_text = pipeline("image-to-text", model="Salesforce/blip-image-captioning-large")
+# image_to_text = pipeline("image-to-text", model="Salesforce/blip-image-captioning-large")
 
 
 # Set OpenAI API key
@@ -40,7 +49,7 @@ def split_image(image, output_dir='.'):
         part_img.save(file_path)
         
         # 生成图片描述
-        caption = image_to_text(part_img)
+        caption = st.session_state.model(part_img)
         results.append((file_path, caption))
     
     return results
@@ -93,6 +102,9 @@ It rains dogs and cats when they were going to the plane. “What a bad weather!
 One days ago, they arrived at there on Saturday. 
 A local person, who is the photographer, helped they took a photo. They had a happy holiday.
 
+image content: a cartoon of a family sitting on a couch reading a book，a cartoon drawing of a group of people climbing up a flight of stairs
+，a cartoon drawing of a family standing on the beach
+
 you should return:
 
 Total mark: 15 points
@@ -137,45 +149,51 @@ Language Dimension 3: Accuracy
 """
 
 example_2 = """
-student essay:In the first picture the family is looking for a maps and magazine of Hawai. In the second picture is raining and the family is took a plane. In the thirt picture the family is in the beach and have sunglasses.
+student essay:
+There is a Jack’s birthday. There are he, his brother, mum and dad. His brother is giving him a ball to football. Next day, Jack and his brother are playing football match. Jack score a goal.
+
+image content: cartoon of a family sitting around a table with a birthday cake，cartoon of a man giving a birthday cake to a woman，
+cartoon of a man kicking a soccer ball into a goal
+
 you should return:
 
-Total mark: 6 points
+Total mark: 9 points
 ------------------------------------------------------------------------------------
 Content 
 
-3 points
+3 points:
 
 - Each picture involves content: ✔️
-- Each image should have 2-3 sentences; if sub-clauses are used, description can be reduced to 1-2 sentences for each image: ❌
-- Writing is partially relevant and some information is not sufficiently conveyed: -
+- Each picture has at least one sentence: ✔️
+- The content is relevant, but the information derived from the image is not exhaustive: ❌
 
-------------------------------------------------------------------------------------
 Organization
 
-1 point
+3 points:
 
-- Structured writing with a clear introduction, body, and conclusion. The first two sentences provide clear context regarding time, place, individuals, and actions: ✔️
-- Less than 90% logical connection between sentences: ❌
-- Insufficient use of linkage techniques: ❌
+- Complete structure with a clear introduction, development, and conclusion: ✔️
+- 50% of the sentences have logical connections: ✔️
+- Contains one to two types of cohesive devices, such as pronouns (e.g., he, his brother) and temporal relations (e.g., next day). The number of cohesive devices used ranges between 2-4 for one type: ✔️
 
-------------------------------------------------------------------------------------
 Language 
 
-2 points
+3 points:
 
 *Dimension 1: Complexity*
-- About 30% of words are bi-syllabic or multi-syllabic: ✔️ 
-- Lacks B1 level or above vocabulary: ❌
-- Does not incorporate sub-clauses: ❌
-- Only one coordinated sentence using "and": -
+
+- Bi-syllabic and multi-syllabic words make up 9%: ✔️
+- All words are common, everyday vocabulary: ✔️
+- No complex sentences or compound sentences: ❌
 
 *Dimension 2: Diversity*
-- Repetitive use of words such as "picture" (3 times), "family" (3 times), and present continuous tense (2 times): ❌
+
+- Repetition of words: "Brother" appears 3 times, "football" appears 2 times: ❌
 
 *Dimension 3: Accuracy*
-- Two word spelling errors: "Hawai", "thirt". Grammatical errors in 2 places: "is took a plane", "is have sunglasses": ❌
-- Can form complete sentences and uses basic words, so the examiner might award an additional 1 point for effort.
+
+- No word spelling mistakes: ✔️
+- Did not use past tense: ❌
+- Grammatical errors in four places: "There is a Jack’s birthday", "There are he", "to football", "Jack score": ❌
 """
 
 system_message = f'''
@@ -258,9 +276,9 @@ with col1:
             st.image(uploaded_image, caption="已上传的图片。", use_column_width=True)
             image = Image.open(uploaded_image).convert("RGB") 
             output_dir = '.'
-            caption_results = split_image(image, output_dir)
+            st.session_state.result = split_image(image, output_dir)
 
-    student_essay = st.text_area(
+    st.session_state.essay = st.text_area(
     " 输入你的作文",
     height= 400
     )
@@ -285,13 +303,13 @@ with col2:
         progress_placeholder = st.empty()
         progress_text = "开始打分"
         my_bar = progress_placeholder.progress(0, text=progress_text)
-        query = str(student_essay)
+        query = str(st.session_state.essay)
         user_message = query
         messages =  [
         {'role':'system',
             'content': system_message},
         {'role':'user',
-            'content': f"请记住你的设定和评分标准【{query}】, 三张图片对应的captions分别是：{caption_results}"},]
+            'content': f"remember your settings student essay:【{query}】, image content：{st.session_state.result}"},]
         response = get_completion_from_messages(messages)
         my_bar.progress(50, text="开始提供相关素材")
         system_message_4 = '''