2 recommendation gallery

app.py

@@ -349,11 +349,12 @@ def extract_features_siglip(image):
 @spaces.GPU
 def infer(crop_image_path,full_image_path,state,language,task_type=None):
     print("task type",task_type)
-    gallery_output = []
+    style_gallery_output = []
+    item_gallery_output=[]
 
     if task_type=="task 1":
-        gallery_output.append("recomendation_pic/1.8.jpg")
-        gallery_output.append("recomendation_pic/1.9.jpg")
+        item_gallery_output.append("recomendation_pic/1.8.jpg")
+        item_gallery_output.append("recomendation_pic/1.9.jpg")
         input_image = Image.open(full_image_path).convert("RGB")
         input_features = extract_features_siglip(input_image.convert("RGB"))
         input_features = input_features.detach().cpu().numpy()
@@ -364,17 +365,17 @@ def infer(crop_image_path,full_image_path,state,language,task_type=None):
             sim = -distances[0][i]
             image_url = df.iloc[v]["Link"]
             img_retrieved = read_image_from_url(image_url)
-            gallery_output.append(img_retrieved)
+            style_gallery_output.append(img_retrieved)
         if language=="English":
             msg="🖼️ Please refer to the section below to see the recommended results."
         else:
             msg="🖼️  请到下方查看推荐结果。"
         state+=[(None,msg)]
 
-        return gallery_output,state,state
+        return item_gallery_output, style_gallery_output,state,state
     elif task_type=="task 2":
-        gallery_output.append("recomendation_pic/2.8.jpg")
-        gallery_output.append("recomendation_pic/2.9.png")
+        item_gallery_output.append("recomendation_pic/2.8.jpg")
+        item_gallery_output.append("recomendation_pic/2.9.png")
         input_image = Image.open(full_image_path).convert("RGB")
         input_features = extract_features_siglip(input_image.convert("RGB"))
         input_features = input_features.detach().cpu().numpy()
@@ -385,18 +386,18 @@ def infer(crop_image_path,full_image_path,state,language,task_type=None):
             sim = -distances[0][i]
             image_url = df.iloc[v]["Link"]
             img_retrieved = read_image_from_url(image_url)
-            gallery_output.append(img_retrieved)
+            style_gallery_output.append(img_retrieved)
         if language=="English":
             msg="🖼️ Please refer to the section below to see the recommended results."
         else:
             msg="🖼️  请到下方查看推荐结果。"
         state+=[(None,msg)]
 
-        return gallery_output,state,state
+        return item_gallery_output, style_gallery_output,state,state
     
     elif task_type=="task 3":
-        gallery_output.append("recomendation_pic/3.8.png")
-        gallery_output.append("recomendation_pic/3.9.png")
+        item_gallery_output.append("recomendation_pic/3.8.png")
+        item_gallery_output.append("recomendation_pic/3.9.png")
         input_image = Image.open(full_image_path).convert("RGB")
         input_features = extract_features_siglip(input_image.convert("RGB"))
         input_features = input_features.detach().cpu().numpy()
@@ -407,14 +408,14 @@ def infer(crop_image_path,full_image_path,state,language,task_type=None):
             sim = -distances[0][i]
             image_url = df.iloc[v]["Link"]
             img_retrieved = read_image_from_url(image_url)
-            gallery_output.append(img_retrieved)
+            style_gallery_output.append(img_retrieved)
         if language=="English":
             msg="🖼️ Please refer to the section below to see the recommended results."
         else:
             msg="🖼️  请到下方查看推荐结果。"
         state+=[(None,msg)]
 
-        return gallery_output,state,state
+        return item_gallery_output, style_gallery_output,state,state
     
     elif crop_image_path:
         input_image = Image.open(crop_image_path).convert("RGB")
@@ -427,7 +428,7 @@ def infer(crop_image_path,full_image_path,state,language,task_type=None):
             sim = -distances[0][i]
             image_url = df.iloc[v]["Link"]
             img_retrieved = read_image_from_url(image_url)
-            gallery_output.append(img_retrieved)
+            item_gallery_output.append(img_retrieved)
 
         input_image = Image.open(full_image_path).convert("RGB")
         input_features = extract_features_siglip(input_image.convert("RGB"))
@@ -439,14 +440,14 @@ def infer(crop_image_path,full_image_path,state,language,task_type=None):
             sim = -distances[0][i]
             image_url = df.iloc[v]["Link"]
             img_retrieved = read_image_from_url(image_url)
-            gallery_output.append(img_retrieved)
+            style_gallery_output.append(img_retrieved)
         if language=="English":
             msg="🖼️ Please refer to the section below to see the recommended results."
         else:
             msg="🖼️  请到下方查看推荐结果。"
         state+=[(None,msg)]
 
-        return gallery_output,state,state
+        return item_gallery_output, style_gallery_output,state,state
     else:
         input_image = Image.open(full_image_path).convert("RGB")
         input_features = extract_features_siglip(input_image.convert("RGB"))
@@ -458,14 +459,14 @@ def infer(crop_image_path,full_image_path,state,language,task_type=None):
             sim = -distances[0][i]
             image_url = df.iloc[v]["Link"]
             img_retrieved = read_image_from_url(image_url)
-            gallery_output.append(img_retrieved)
+            style_gallery_output.append(img_retrieved)
         if language=="English":
             msg="🖼️ Please refer to the section below to see the recommended results."
         else:
             msg="🖼️  请到下方查看推荐结果。"
         state+=[(None,msg)]
 
-        return gallery_output,state,state
+        return item_gallery_output, style_gallery_output,state,state
         
 
 
@@ -1646,10 +1647,10 @@ async def texttospeech(text, language,gender='female'):
         return None
 
 # give the reason of recommendation
-async def associate(image_path,new_crop,openai_api_key,language,autoplay,length,log_state,sort_score,narritive,state,evt: gr.SelectData):
+async def item_associate(image_path,new_crop,openai_api_key,language,autoplay,length,log_state,sort_score,narritive,state,evt: gr.SelectData):
     persona=naritive_mapping[narritive]
     rec_path=evt._data['value']['image']['path']
-    index=evt.index
+    index="Item Recommendation Picture "+str(evt.index)
     print("rec_path",rec_path)
     prompt=recommendation_prompt[persona].format(language=language,length=length)
     if new_crop:
@@ -1670,6 +1671,32 @@ async def associate(image_path,new_crop,openai_api_key,language,autoplay,length,
         audio_output = await texttospeech(read_info, language)
     return state,state,audio_output,log_state,index,gr.update(value=[])
 
+
+async def style_associate(image_path,new_crop,openai_api_key,language,autoplay,length,log_state,sort_score,narritive,state,evt: gr.SelectData):
+    persona=naritive_mapping[narritive]
+    rec_path=evt._data['value']['image']['path']
+    index="Style Recommendation Picture "+str(evt.index)
+    print("rec_path",rec_path)
+    prompt=recommendation_prompt[persona].format(language=language,length=length)
+    if new_crop:
+        image_paths=[new_crop,rec_path]
+    else:
+        image_paths=[image_path,rec_path]
+    result=get_gpt_response(openai_api_key, image_paths, prompt)
+    print("recommend result",result)
+    state += [(None, f"{result}")]
+    log_state = log_state + [(narritive, None)]
+    log_state = log_state + [(f"image sort ranking {sort_score}", None)]
+    log_state = log_state + [(None, f"{result}")]
+    read_info = re.sub(r'[#[\]!*]','',result)
+    read_info = emoji.replace_emoji(read_info,replace="")  
+    print("associate",read_info)
+    audio_output=None
+    if autoplay:
+        audio_output = await texttospeech(read_info, language)
+    return state,state,audio_output,log_state,index,gr.update(value=[])
+
+
 def change_naritive(session_type,image_input, state, click_state, paragraph, origin_image,narritive,task_instruct,gallery_output,reco_reasons,language="English"):
     if session_type=="Session 1":
         return None, [], [], [[], [], []], "", None, None, [], [],[]
@@ -1736,7 +1763,7 @@ def print_like_dislike(x: gr.LikeData,state,log_state):
     return log_state,state
 
 def get_recommendationscore(index,score,log_state):
-    log_state+=[(f"Picture {index} : {score}",None)]
+    log_state+=[(f"{index} : {score}",None)]
     log_state+=[("%% recommendation %%",None)]
     return log_state
     
@@ -2009,9 +2036,24 @@ def create_ui():
                     )
                     
                     gallery_result = gr.Gallery(
-                    label="Recommendations",
+                    label="Recommendations Based on Item",
                     height="auto",
-                    columns=4
+                    columns=2
+                    # columns=4, 
+                    # rows=2,
+                    # show_label=False, 
+                    # allow_preview=True, 
+                    # object_fit="contain", 
+                    # height="auto", 
+                    # preview=True, 
+                    # show_share_button=True, 
+                    # show_download_button=True
+                    )
+                    
+                    style_gallery_result = gr.Gallery(
+                    label="Recommendations Based on Style",
+                    height="auto",
+                    columns=2
                     # columns=4, 
                     # rows=2,
                     # show_label=False, 
@@ -2208,11 +2250,19 @@ def create_ui():
         recommend_btn.click(
             fn=infer,
             inputs=[new_crop_save_path,image_path,state,language,task_type],
-            outputs=[gallery_result,chatbot,state]
+            outputs=[gallery_result,style_gallery_result,chatbot,state]
             )
 
         gallery_result.select(
-            associate,
+            item_associate,
+            inputs=[image_path,new_crop_save_path,openai_api_key,language,auto_play,length,log_state,sort_rec,naritive,recomended_state],
+            outputs=[recommend_bot,recomended_state,output_audio,log_state,pic_index,recommend_score],
+            
+            
+        )
+        
+        style_gallery_result.select(
+            style_associate,
             inputs=[image_path,new_crop_save_path,openai_api_key,language,auto_play,length,log_state,sort_rec,naritive,recomended_state],
             outputs=[recommend_bot,recomended_state,output_audio,log_state,pic_index,recommend_score],