mid conv result

@@ -11,7 +11,7 @@ def softmax(inputA):
     normalized_arr = []
     for x in inputA:
         normalized_arr.append(float(x))
-    #normalized_arr = normalize(normalized_arr) 
+    normalized_arr = normalize(normalized_arr) 
         
     for i in range(0, len(normalized_arr)):
         
@@ -36,5 +36,5 @@ def normalize(arrs):
     minimum = min(normalized_arr)
     denom = float(maximum) - float(minimum)
     for i in range(0,len(normalized_arr)):
-        normalized_arr[i] = (normalized_arr[i] - minimum)/ denom
+        normalized_arr[i] = ((normalized_arr[i] - minimum)/ denom) * 2 - 1
     return normalized_arr
\ No newline at end of file

@@ -10,11 +10,11 @@ import video_util as videoutil
 
 # Define file paths.
 MODEL_PATH = "/mnt/e/khuhub/2015104192/web/backend/yt8m/esot3ria/model/inference_model/segment_inference_model"
-VOCAB_PATH = "/mnt/e/khuhub/2015104192/web/backend/yt8m/prevs/vocabulary.csv"
+VOCAB_PATH = "/mnt/e/khuhub/2015104192/web/backend/yt8m/vocabulary.csv"
 VIDEO_TAGS_PATH = "/mnt/e/khuhub/2015104192/web/backend/yt8m/esot3ria/segment_tags.csv"
 TAG_VECTOR_MODEL_PATH = "/mnt/e/khuhub/2015104192/web/backend/yt8m/esot3ria/tag_vectors.model"
 VIDEO_VECTOR_MODEL_PATH = "/mnt/e/khuhub/2015104192/web/backend/yt8m/esot3ria/video_vectors.model"
-SEGMENT_LABEL_PATH = "/mnt/e/khuhub/2015104192/web/backend/yt8m//prevs/segment_label_ids.csv"
+SEGMENT_LABEL_PATH = "/mnt/e/khuhub/2015104192/web/backend/yt8m/segment_label_ids.csv"
  
 # Define parameters.
 TAG_TOP_K = 5

@@ -2,7 +2,7 @@ import nltk
 import gensim
 import pandas as pd
 from gensim.models import Word2Vec
-
+import sys
 def normalize_tag(tag):
     if isinstance(tag, str):
         new_tag = tag.lower().replace('[^a-zA-Z]', ' ')
@@ -13,7 +13,7 @@ def normalize_tag(tag):
         return new_tag
     else:
         return tag
-
+'''
 # Load files.
 nltk.download('stopwords')
 vocab = pd.read_csv('E:/khuhub/2015104192/web/backend/yt8m/esot3ria/vocabulary.csv',encoding='utf-8')
@@ -49,11 +49,14 @@ vocab_phrased = phraser[tokenlist]
 
 # Vectorize tags.
 w2v = gensim.models.word2vec.Word2Vec(sentences=vocab_phrased, min_count=1)
-w2v.save('E:/khuhub/2015104192/web/backend/yt8m/esot3ria/tag_vectors3.model')
-
-
-tag_vectors = Word2Vec.load("./tag_vectors3.model").wv
-print(tag_vectors['concert'])
+w2v.save('E:/khuhub/2015104192/web/backend/yt8m/esot3ria/tag_vectors.model')
+'''
+
+tag_vectors = Word2Vec.load("./tag_vectors.model").wv
+print(tag_vectors.similarity('koi','koi'))
+all_sims = tag_vectors.most_similar('koi', topn=sys.maxsize)
+last_10 = list(reversed(all_sims[-10:]))
+print(last_10)
 # word_vectors = w2v.wv
 # vocabs = word_vectors.vocab.keys()
 # word_vectors_list = [word_vectors[v] for v in vocabs]
\ No newline at end of file

@@ -22,15 +22,16 @@ def recommend_videos(tags, segments, tag_model_path, video_model_path, top_k):
     maxSimilarity = -1
     
     kernel = [np.zeros(100) for i in range(0,5)]
-
+    tagKernel = []
     #우선은 비교를 뜰 입력 영상의 단일 비디오벡터를 구함
     video_vector = np.zeros(100)
     tag_preds =[]
-    
+    videoTagList = []
     for (tag, weight) in tags:
         tag_preds.append(weight)
+        videoTagList.append(tag)
     #print("tag preds = ",tag_preds) 
-    tag_preds = ac.softmax(tag_preds)
+    #tag_preds = ac.softmax(tag_preds)
     for (tag, weight),pred in zip(tags,tag_preds):
         print(tag,pred)
         if tag in tag_vectors.vocab:
@@ -47,7 +48,7 @@ def recommend_videos(tags, segments, tag_model_path, video_model_path, top_k):
     for segment in segments:
         segment_vector = np.zeros(100)
         segTags = [segment[i] for i in range(0,len(segment),2)]
-        segProbs = ac.softmax([float(segment[i]) for i in range(1,len(segment),2)])
+        segProbs = [float(segment[i]) for i in range(1,len(segment),2)]#ac.softmax([float(segment[i]) for i in range(1,len(segment),2)])
         
         #print(segProbs)
         for tag, weight in zip(segTags,segProbs):
@@ -80,11 +81,11 @@ def recommend_videos(tags, segments, tag_model_path, video_model_path, top_k):
         segment = segments[maxSimilarSegment - math.floor(len(kernel)/2) + k]
         segment_vector = np.zeros(100)
         segTags = [segment[i] for i in range(0,len(segment),2)]
-        
+        tagKernel.append(segTags)
         segProbs = ac.softmax([float(segment[i]) for i in range(1,len(segment),2)])
         print(segTags)
         print(segProbs)
-        normalize(segProbs)
+        #normalize(segProbs)
         for (tag, weight) in zip(segTags,segProbs):
             if tag in tag_vectors.vocab:
                 #float(weight)
@@ -96,11 +97,22 @@ def recommend_videos(tags, segments, tag_model_path, video_model_path, top_k):
                     error_tags.append(tag)
                    
         kernel[k] = segment_vector
-
+        '''
+        if(k < int(len(kernel)/2)):
+            kernel[k] = kernel[k] * k
+        elif(k > int(len(kernel)/2)):
+            kernel[k] = kernel[k] * (len(kernel) - k)
+        else:
+            kernel[k] = kernel[k] * (len(kernel)/2 + 1)
+        '''
+    print("TAG kernel")
+    #tagKernel = tagKernel[1:5] 
+    print(tagKernel)
     #여기에서 유사한 벡터들을 뽑아냄
     #현재는 비디오id로 영상을 얻을 수 없으므로 반환값으로 비디오 아이디와 태그들, 확률 사용
     video_tags = pd.read_csv('/mnt/e/khuhub/2015104192/web/backend/yt8m/esot3ria/segment_tags.csv', encoding='utf8',error_bad_lines=False)
     videoVectorList = []
+    segmentTagList = []
     prevVideoId = ""
     minimunVideoIds = [["",-1.0] for i in range(0,top_k)]
 
@@ -117,7 +129,10 @@ def recommend_videos(tags, segments, tag_model_path, video_model_path, top_k):
                 #print("====")
                 #for a in kernel:
                 #    print(len(kernel),norm(a))
-            maxima, idx = convolution(videoVectorList,kernel,prevVideoId)
+            convmaxima, convidx = convolution(videoVectorList,kernel,prevVideoId)
+            maxima, idx = differenceMax(segmentTagList,tagKernel,tag_vectors,videoTagList)
+            #maxima = maxima + convmaxima
+            
             #print(video_id,maxima)
             localMinima = 100
             localMinimaIndex = -1
@@ -127,16 +142,21 @@ def recommend_videos(tags, segments, tag_model_path, video_model_path, top_k):
                     localMinimaIndex = seg  
             #print(maxima)
             if localMinima < maxima:
-                print(prevVideoId[0:4] + "_" + str(idx),maxima)
+                print(prevVideoId[0:4] + "_" + str(idx),localMinimaIndex,maxima,convmaxima)
                 minimunVideoIds[localMinimaIndex] = [prevVideoId[0:4] + "_" + str(idx),maxima]
 
             
             videoVectorList.clear()
+            segmentTagList.clear()
             prevVideoId = video_id
         
         if video_id == "finished":
             break
         videoVectorList.append(video_vectors[video_id])
+        tagList = []
+        for i in range(1,top_k+1):
+            tagList.append(row[i].split(":")[0])
+        segmentTagList.append(tagList)
     
     similar_ids = []
     for i in range(0,top_k):
@@ -155,30 +175,65 @@ def cos_sim(A, B):
         return 0
     else:
         return  dot(A, B)/(norm(A)*norm(B))
-
-def shiftKernel(kernel, newValue):
-    for i in range(0, len(kernel) - 1):
-        kernel[i] = kernel[i+1]
-    kernel[len(kernel) - 1] = newValue
     
+def sub_vec_size(A,B):
+    dir = A-B
+    return norm(dir)
+
 def convolution(arrs, _kernel,vidId):
     s = len(_kernel)
-    
+    l = len(arrs)
     result = []
     
     midpos = math.floor(s/2)
     for i in range(0,midpos):
         arrs.insert(0,np.zeros(100))
         arrs.append(np.zeros(100))
-        
+
+    total = 0
     for j in range(midpos,len(arrs) - midpos):
         convResult = 0
         for i in range(0, s):
-            convResult = convResult + cos_sim(arrs[j - midpos + i],_kernel[i])
+            if(i == int(len(_kernel)/2)):
+                convResult = convResult - sub_vec_size(arrs[j - midpos + i],_kernel[i]) + dot(arrs[j - midpos + i],_kernel[i])
         result.append(convResult)
+        total = total + convResult
     maxVal = max(result)    
     index = result.index(maxVal)
-    return maxVal,index
+    
+    return total/l,index
+
+def differenceMax(arrs, _kernel, w2v, videoTaglist):
+    s = len(_kernel)
+    
+    result = []
+    
+    midpos = math.floor(s/2)
+    for i in range(0,midpos):
+        arrs.insert(0,arrs[0])
+        arrs.append(arrs[len(arrs)-1])
+
+    prevIndex = 0
+    prevMax = -100
+    for j in range(midpos,len(arrs) - midpos):
+        convResult = 0 
+        processed_vocabNum = 1
+        for i in range(0, s):
+            #if i == midpos:
+            if(_kernel[i][0] not in arrs[j - midpos + i][0:2]):# and ((videoTaglist[0] not in arrs[j - midpos + i][0:2])) and ((videoTaglist[1] not in arrs[j - midpos + i][0:5])):
+                continue
+            for ind in range(0,5):
+                if(arrs[j - midpos + i][ind] in w2v.vocab) and (_kernel[i][ind] in w2v.vocab):
+                    convResult = convResult + (w2v.similarity(arrs[j - midpos + i][ind],_kernel[i][ind]))
+                    processed_vocabNum = processed_vocabNum + 1
+        #convResult = convResult / processed_vocabNum
+        if prevMax < convResult:
+            prevMax = convResult
+            prevIndex = j - midpos
+        result.append(convResult)
+    #maxVal = max(result)    
+    #index = result.index(maxVal)
+    return prevMax,prevIndex
 
 def normalize(arrs):    
     maximum = max(arrs)

@@ -29,7 +29,10 @@ def getVideoInfo(vid_id, video_tags_path, top_k):
         video_tag_tuple = video_tags_info["segment" + str(i)].values[0]# ex: "mobile-phone:0.361"
         video_tags.append(video_tag_tuple.split(":")[0])
         if video_url == "":
-            video_url = video_url + ' ' + video_tags
+            for x in video_tags:
+                video_url = video_url + ' ' + x
+                
+    video_url = video_url + '\nThe similar point is : ' + str(float(vid_id[5:]) * 5)
 
     return {
         "video_url": video_url,