made it run

++Subproject commit 8ed8b52992cd8b25cbc8a25c1513ed0a0809de87

@@ -18,8 +18,9 @@ import shlex
 import json
 import json
 # Create your views here.
 # Create your views here.
 import sys
 import sys
-sys.path.insert(0, "/khuhub/2015104192/web/backend/yt8m/esot3ria")
+sys.path.insert(0, "/mnt/e/khuhub/2015104192/web/backend/yt8m/esot3ria")
 import inference_pb
 import inference_pb
+import pb_util as pbutil
 
 
 def with_ffprobe(filename):
 def with_ffprobe(filename):
 
 
@@ -77,16 +78,16 @@ class VideoFileUploadView(APIView):
             file_serializer.save()     
             file_serializer.save()     
             # 동영상 길이 출력  
             # 동영상 길이 출력  
             fileFullName = file_serializer.data['file_save_name']
             fileFullName = file_serializer.data['file_save_name']
-            fileProcessedName = fileFullName.replace("%3A",":")
+            print('FILE======================================================= ' + fileFullName)
+            fileProcessedName = fileFullName.replace("E%3A","")
+            print('FILE======================================================= ' + fileProcessedName)
             runTime = with_ffprobe(fileProcessedName)
             runTime = with_ffprobe(fileProcessedName)
-            print('RUNTIME ::::==== ' + str(runTime))
             print(threshold)
             print(threshold)
-            print('CURR DIR = ' + os.getcwd())
-            process = subprocess.Popen(['./runMediaPipe.sh %s %s' %(fileProcessedName,runTime,)], shell = True)
+            print('runtime = ' + str(runTime))
+            process = subprocess.Popen(["sh runMediaPipe.sh %s %d " %(fileProcessedName,runTime)], shell = True)
             process.wait()
             process.wait()
-
-
-            result =  inference_pb.inference_pb('E:/khuhub/2015104192/web/backend/yt8m/featuremaps/features(baby_samoyed).pb', threshold)
+            print("NOW DONE name = " + fileProcessedName + " / thresh = " + str(threshold))
+            result =  inference_pb.inference_pb("/tmp/mediapipe/features.pb", threshold)
 
 
             return Response(result, status=status.HTTP_201_CREATED)
             return Response(result, status=status.HTTP_201_CREATED)
         else:
         else:

+#!/bin bash
+cd ../../mediapipe
+echo "${PWD}"
\ No newline at end of file

 #!/bin/bash
 #!/bin/bash
-cd ../../../../mediapipe
+echo "${PWD}"
+cd ../../mediapipe/mediapipe
 . venv/bin/activate
 . venv/bin/activate
-echo "STEP1"
-bazel version && \
-alias bazel='bazel'
-
-echo "STEP2"
-
+echo "activated"
+/usr/local/bazel/3.4.0/lib/bazel/bin/bazel version && \
+alias bazel='/usr/local/bazel/3.4.0/lib/bazel/bin/bazel'
 python -m mediapipe.examples.desktop.youtube8m.generate_input_sequence_example \
 python -m mediapipe.examples.desktop.youtube8m.generate_input_sequence_example \
   --path_to_input_video=/$1 \
   --path_to_input_video=/$1 \
   --clip_end_time_sec=$2
   --clip_end_time_sec=$2
-echo "STEP3"
-GLOG_logtostderr=1 /mediapipe/examples/desktop/youtube8m/extract_yt8m_features \
+echo "Nocly done"
+GLOG_logtostderr=1 bazel-bin/mediapipe/examples/desktop/youtube8m/extract_yt8m_features \
   --calculator_graph_config_file=mediapipe/graphs/youtube8m/feature_extraction.pbtxt \
   --calculator_graph_config_file=mediapipe/graphs/youtube8m/feature_extraction.pbtxt \
   --input_side_packets=input_sequence_example=/tmp/mediapipe/metadata.pb  \
   --input_side_packets=input_sequence_example=/tmp/mediapipe/metadata.pb  \
-  --output_side_packets=output_sequence_example=/tmp/mediapipe/features.pb
-echo "COMPLETED"
-
-sleep 5
\ No newline at end of file
+  --output_side_packets=output_sequence_example=/tmp/mediapipe/features.pb
\ No newline at end of file

+#!/bin bash
+
+cd ../../../mediapipe
+. venv/bin/activate
+
+/usr/local/bazel/3.4.0/lib/bazel/bin/bazel version && \
+alias bazel='/usr/local/bazel/3.4.0/lib/bazel/bin/bazel'
+
+python -m mediapipe.examples.desktop.youtube8m.generate_input_sequence_example \
+  --path_to_input_video=/$1 \
+  --clip_end_time_sec=$2
+
+GLOG_logtostderr=1 bazel-bin/mediapipe/examples/desktop/youtube8m/extract_yt8m_features \
+  --calculator_graph_config_file=mediapipe/graphs/youtube8m/feature_extraction.pbtxt \
+  --input_side_packets=input_sequence_example=/tmp/mediapipe/metadata.pb  \
+  --output_side_packets=output_sequence_example=/tmp/mediapipe/features.pb
\ No newline at end of file

+{"model": "FrameLevelLogisticModel", "feature_sizes": "1024,128", "feature_names": "rgb,audio", "frame_features": true, "label_loss": "CrossEntropyLoss"}
\ No newline at end of file

@@ -9,12 +9,12 @@ import video_recommender as recommender
 import video_util as videoutil
 import video_util as videoutil
 
 
 # Define file paths.
 # Define file paths.
-MODEL_PATH = "E:/khuhub/2015104192/web/backend/yt8m/esot3ria/model/inference_model/segment_inference_model"
-VOCAB_PATH = "E:/khuhub/2015104192/web/backend/yt8m/vocabulary.csv"
-VIDEO_TAGS_PATH = "E:/khuhub/2015104192/web/backend/yt8m/esot3ria/kaggle_solution_40k.csv"
-TAG_VECTOR_MODEL_PATH = "E:/khuhub/2015104192/web/backend/yt8m/esot3ria/tag_vectors.model"
-VIDEO_VECTOR_MODEL_PATH = "E:/khuhub/2015104192/web/backend/yt8m/esot3ria/video_vectors.model"
-SEGMENT_LABEL_PATH = "E:/khuhub/2015104192/web/backend/yt8m/segment_label_ids.csv"
+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/vocabulary.csv"
+VIDEO_TAGS_PATH = "/mnt/e/khuhub/2015104192/web/backend/yt8m/esot3ria/kaggle_solution_40k.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/segment_label_ids.csv"
 
 
 # Define parameters.
 # Define parameters.
 TAG_TOP_K = 5
 TAG_TOP_K = 5
@@ -89,11 +89,12 @@ def inference_pb(file_path, threshold):
     print("started step 1")
     print("started step 1")
     VIDEO_TOP_K = int(threshold)
     VIDEO_TOP_K = int(threshold)
     inference_result = {}
     inference_result = {}
+    print("STEP ==== 1")
     with tf.Session(config=tf.ConfigProto(allow_soft_placement=True)) as sess:
     with tf.Session(config=tf.ConfigProto(allow_soft_placement=True)) as sess:
-
+        print("STEP ==== 1.5 " + file_path)
         # 0. Import SequenceExample type target from pb.
         # 0. Import SequenceExample type target from pb.
         target_video = pbutil.convert_pb(file_path)
         target_video = pbutil.convert_pb(file_path)
-
+        print("STEP ==== 2")
         # 1. Load video features from pb.
         # 1. Load video features from pb.
         video_id_batch_val = np.array([b'video'])
         video_id_batch_val = np.array([b'video'])
         n_frames = len(target_video.feature_lists.feature_list['rgb'].feature)
         n_frames = len(target_video.feature_lists.feature_list['rgb'].feature)
@@ -109,13 +110,13 @@ def inference_pb(file_path, threshold):
             video_batch_val[i] = np.concatenate([video_batch_rgb, video_batch_audio], axis=0)
             video_batch_val[i] = np.concatenate([video_batch_rgb, video_batch_audio], axis=0)
         video_batch_val = np.array([video_batch_val])
         video_batch_val = np.array([video_batch_val])
         num_frames_batch_val = np.array([n_frames])
         num_frames_batch_val = np.array([n_frames])
-
+        print("STEP ==== 3")
         # Restore checkpoint and meta-graph file.
         # Restore checkpoint and meta-graph file.
         if not gfile.Exists(MODEL_PATH + ".meta"):
         if not gfile.Exists(MODEL_PATH + ".meta"):
             raise IOError("Cannot find %s. Did you run eval.py?" % MODEL_PATH)
             raise IOError("Cannot find %s. Did you run eval.py?" % MODEL_PATH)
         meta_graph_location = MODEL_PATH + ".meta"
         meta_graph_location = MODEL_PATH + ".meta"
         logging.info("loading meta-graph: " + meta_graph_location)
         logging.info("loading meta-graph: " + meta_graph_location)
-
+        print("STEP ==== 4")
         with tf.device("/cpu:0"):
         with tf.device("/cpu:0"):
             saver = tf.train.import_meta_graph(meta_graph_location, clear_devices=True)
             saver = tf.train.import_meta_graph(meta_graph_location, clear_devices=True)
         logging.info("restoring variables from " + MODEL_PATH)
         logging.info("restoring variables from " + MODEL_PATH)
@@ -123,7 +124,7 @@ def inference_pb(file_path, threshold):
         input_tensor = tf.get_collection("input_batch_raw")[0]
         input_tensor = tf.get_collection("input_batch_raw")[0]
         num_frames_tensor = tf.get_collection("num_frames")[0]
         num_frames_tensor = tf.get_collection("num_frames")[0]
         predictions_tensor = tf.get_collection("predictions")[0]
         predictions_tensor = tf.get_collection("predictions")[0]
-
+        print("STEP ==== 5")
         # Workaround for num_epochs issue.
         # Workaround for num_epochs issue.
         def set_up_init_ops(variables):
         def set_up_init_ops(variables):
             init_op_list = []
             init_op_list = []
@@ -133,10 +134,10 @@ def inference_pb(file_path, threshold):
                     variables.remove(variable)
                     variables.remove(variable)
             init_op_list.append(tf.variables_initializer(variables))
             init_op_list.append(tf.variables_initializer(variables))
             return init_op_list
             return init_op_list
-
+        print("STEP ==== 6")
         sess.run(
         sess.run(
             set_up_init_ops(tf.get_collection_ref(tf.GraphKeys.LOCAL_VARIABLES)))
             set_up_init_ops(tf.get_collection_ref(tf.GraphKeys.LOCAL_VARIABLES)))
-
+        print("STEP ==== 7")
         whitelisted_cls_mask = np.zeros((predictions_tensor.get_shape()[-1],),
         whitelisted_cls_mask = np.zeros((predictions_tensor.get_shape()[-1],),
                                         dtype=np.float32)
                                         dtype=np.float32)
         with tf.io.gfile.GFile(SEGMENT_LABEL_PATH) as fobj:
         with tf.io.gfile.GFile(SEGMENT_LABEL_PATH) as fobj:
@@ -221,5 +222,5 @@ def inference_pb(file_path, threshold):
 
 
 if __name__ == '__main__':
 if __name__ == '__main__':
     filepath = "/tmp/mediapipe/features.pb"
     filepath = "/tmp/mediapipe/features.pb"
-    result = inference_pb(filepath)
+    result = inference_pb(filepath,5)
     print(result)
     print(result)

 import tensorflow as tf
 import tensorflow as tf
 import numpy
 import numpy
+import pickle
 
 
 
 
 def _make_bytes(int_array):
 def _make_bytes(int_array):
@@ -43,17 +44,20 @@ features = {
 
 
 
 
 def parse_exmp(serial_exmp):
 def parse_exmp(serial_exmp):
+    print("CAME IN SECOND======111111 : ")
     _, sequence_parsed = tf.io.parse_single_sequence_example(
     _, sequence_parsed = tf.io.parse_single_sequence_example(
         serialized=serial_exmp,
         serialized=serial_exmp,
         context_features=contexts,
         context_features=contexts,
         sequence_features=features)
         sequence_features=features)
-
-    sequence_parsed = tf.contrib.learn.run_n(sequence_parsed)[0]
-
+    print("CAME IN SECOND=====222222 := ")
+    sequence_parsed = tf.contrib.learn.run_n(sequence_parsed, n=1, feed_dict=None)[0]
+    print("CAME IN SECOND=====3333333")
     audio = sequence_parsed['AUDIO/feature/floats'].values
     audio = sequence_parsed['AUDIO/feature/floats'].values
+    print("CAME IN SECOND=====44444444")
     rgb = sequence_parsed['RGB/feature/floats'].values
     rgb = sequence_parsed['RGB/feature/floats'].values
-
-    # print(audio.values)
+    print("CAME IN SECOND=====5555555")
+    print("======AUDIO=======")
+    print(audio)
     # print(type(audio.values))
     # print(type(audio.values))
 
 
     # audio is 128 8bit, rgb is 1024 8bit for every second
     # audio is 128 8bit, rgb is 1024 8bit for every second
@@ -112,9 +116,21 @@ def make_exmp(id, audio, rgb):
 
 
 
 
 def convert_pb(filename):
 def convert_pb(filename):
+    print("==============")
+    print("CAME IN! " + filename)
+    print("==============")
     sequence_example = open(filename, 'rb').read()
     sequence_example = open(filename, 'rb').read()
+    print("==============")
+    print("IM 1 :::    ")
+    print("==============")
     audio, rgb = parse_exmp(sequence_example)
     audio, rgb = parse_exmp(sequence_example)
+    print("==============")
+    print("IM 2")
+    print("==============")
     tmp_example = make_exmp('video', audio, rgb)
     tmp_example = make_exmp('video', audio, rgb)
-
+    print("==============")
+    print("IM 3")
+    print("==============")
     decoded = tf.train.SequenceExample.FromString(tmp_example)
     decoded = tf.train.SequenceExample.FromString(tmp_example)
+    print("DECODE COMPLETE")
     return decoded
     return decoded

@@ -12,6 +12,8 @@ def getURL(vid_id):
         return youtube_url + response.text[10:-3]
         return youtube_url + response.text[10:-3]
 
 
 
 
+print(getURL('udtT'))
+
 def getVideoInfo(vid_id, video_tags_path, top_k):
 def getVideoInfo(vid_id, video_tags_path, top_k):
     video_url = getURL(vid_id)
     video_url = getURL(vid_id)
 
 

@@ -33,7 +33,7 @@ FLAGS = flags.FLAGS
 if __name__ == "__main__":
 if __name__ == "__main__":
   # Dataset flags.
   # Dataset flags.
   flags.DEFINE_string(
   flags.DEFINE_string(
-      "train_dir", "/tmp/yt8m_model/",
+      "train_dir", "F:/yt8mDataset/savedModel",
       "The directory to load the model files from. "
       "The directory to load the model files from. "
       "The tensorboard metrics files are also saved to this "
       "The tensorboard metrics files are also saved to this "
       "directory.")
       "directory.")

@@ -85,7 +85,7 @@ if __name__ == "__main__":
   flags.DEFINE_integer("top_k", 5, "How many predictions to output per video.")
   flags.DEFINE_integer("top_k", 5, "How many predictions to output per video.")
 
 
   # Other flags.
   # Other flags.
-  flags.DEFINE_integer("batch_size", 512,
+  flags.DEFINE_integer("batch_size", 32,
                        "How many examples to process per batch.")
                        "How many examples to process per batch.")
   flags.DEFINE_integer("num_readers", 1,
   flags.DEFINE_integer("num_readers", 1,
                        "How many threads to use for reading input files.")
                        "How many threads to use for reading input files.")
@@ -276,7 +276,7 @@ def inference(reader, train_dir, data_pattern, out_file_location, batch_size,
     #open vocab csv file and store to dictionary
     #open vocab csv file and store to dictionary
     #=========================================
     #=========================================
     voca_dict = {}
     voca_dict = {}
-    vocabs = open("./vocabulary.csv", 'r')
+    vocabs = open("./vocabulary.csv", 'r',encoding='UTF8')
     while True:
     while True:
       line = vocabs.readline()
       line = vocabs.readline()
       if not line: break
       if not line: break

@@ -75,7 +75,7 @@ if __name__ == "__main__":
   flags.DEFINE_integer(
   flags.DEFINE_integer(
       "num_gpu", 1, "The maximum number of GPU devices to use for training. "
       "num_gpu", 1, "The maximum number of GPU devices to use for training. "
       "Flag only applies if GPUs are installed")
       "Flag only applies if GPUs are installed")
-  flags.DEFINE_integer("batch_size", 128,
+  flags.DEFINE_integer("batch_size", 64,
                        "How many examples to process per batch for training.")
                        "How many examples to process per batch for training.")
   flags.DEFINE_string("label_loss", "CrossEntropyLoss",
   flags.DEFINE_string("label_loss", "CrossEntropyLoss",
                       "Which loss function to use for training the model.")
                       "Which loss function to use for training the model.")
@@ -83,14 +83,14 @@ if __name__ == "__main__":
       "regularization_penalty", 1.0,
       "regularization_penalty", 1.0,
       "How much weight to give to the regularization loss (the label loss has "
       "How much weight to give to the regularization loss (the label loss has "
       "a weight of 1).")
       "a weight of 1).")
-  flags.DEFINE_float("base_learning_rate", 0.001,
+  flags.DEFINE_float("base_learning_rate", 0.002,
                      "Which learning rate to start with.")
                      "Which learning rate to start with.")
   flags.DEFINE_float(
   flags.DEFINE_float(
       "learning_rate_decay", 0.8,
       "learning_rate_decay", 0.8,
       "Learning rate decay factor to be applied every "
       "Learning rate decay factor to be applied every "
       "learning_rate_decay_examples.")
       "learning_rate_decay_examples.")
   flags.DEFINE_float(
   flags.DEFINE_float(
-      "learning_rate_decay_examples", 100,
+      "learning_rate_decay_examples", 70,
       "Multiply current learning rate by learning_rate_decay "
       "Multiply current learning rate by learning_rate_decay "
       "every learning_rate_decay_examples.")
       "every learning_rate_decay_examples.")
   flags.DEFINE_integer(
   flags.DEFINE_integer(
@@ -100,7 +100,7 @@ if __name__ == "__main__":
       "max_steps", None,
       "max_steps", None,
       "The maximum number of iterations of the training loop.")
       "The maximum number of iterations of the training loop.")
   flags.DEFINE_integer(
   flags.DEFINE_integer(
-      "export_model_steps", 100,
+      "export_model_steps", 5,
       "The period, in number of steps, with which the model "
       "The period, in number of steps, with which the model "
       "is exported for batch prediction.")
       "is exported for batch prediction.")
 
 

+캡1때 했던 컨셉들은 간단하게 적고 넘어간다
+
+1. 여러가지 모델로 학습을 해보고 제일 좋은거 고르자
+1.5 워드벡터 임베딩에 쓸 말뭉치 좋은거 쓰자
+2. 태그의 weight를 어떻게 조절을 해보자
+
+(3. 성능(시간복잡도) 향상)
\ No newline at end of file