Project import generated by Copybara.

GitOrigin-RevId: 4419aaa472eeb91123d1f8576188166ee0e5ea69

mediapipe/examples/desktop/autoflip/README.md

@@ -11,6 +11,8 @@
 
 2.  Build and run the run_autoflip binary to process a local video.
 
+Note: AutoFlip currently only works with OpenCV 3 . Please verify your OpenCV version beforehand.
+
     ```bash
     bazel build -c opt --define MEDIAPIPE_DISABLE_GPU=1 \
       mediapipe/examples/desktop/autoflip:run_autoflip

mediapipe/examples/desktop/media_sequence/charades_dataset.py

@@ -63,12 +63,15 @@ import random
 import subprocess
 import sys
 import tempfile
-import urllib
 import zipfile
+
 from absl import app
 from absl import flags
 from absl import logging
+from six.moves import range
+from six.moves import urllib
 import tensorflow.compat.v1 as tf
+
 from mediapipe.util.sequence import media_sequence as ms
 
 
@@ -218,7 +221,7 @@ class Charades(object):
       return output_dict
 
     if split not in SPLITS:
-      raise ValueError("Split %s not in %s" % split, str(SPLITS.keys()))
+      raise ValueError("Split %s not in %s" % split, str(list(SPLITS.keys())))
     all_shards = tf.io.gfile.glob(
         os.path.join(self.path_to_data, SPLITS[split][0] + "-*-of-*"))
     random.shuffle(all_shards)
@@ -329,7 +332,7 @@ class Charades(object):
     if sys.version_info >= (3, 0):
       urlretrieve = urllib.request.urlretrieve
     else:
-      urlretrieve = urllib.urlretrieve
+      urlretrieve = urllib.request.urlretrieve
     logging.info("Creating data directory.")
     tf.io.gfile.makedirs(self.path_to_data)
     logging.info("Downloading license.")

mediapipe/examples/desktop/media_sequence/demo_dataset.py

@@ -57,11 +57,12 @@ import random
 import subprocess
 import sys
 import tempfile
-import urllib
 
 from absl import app
 from absl import flags
 from absl import logging
+from six.moves import range
+from six.moves import urllib
 import tensorflow.compat.v1 as tf
 
 from mediapipe.util.sequence import media_sequence as ms
@@ -198,7 +199,7 @@ class DemoDataset(object):
     if sys.version_info >= (3, 0):
       urlretrieve = urllib.request.urlretrieve
     else:
-      urlretrieve = urllib.urlretrieve
+      urlretrieve = urllib.request.urlretrieve
     for split in SPLITS:
       reader = csv.DictReader(SPLITS[split].split("\n"))
       all_metadata = []

mediapipe/examples/desktop/media_sequence/kinetics_dataset.py

@@ -73,11 +73,13 @@ import subprocess
 import sys
 import tarfile
 import tempfile
-import urllib
 
 from absl import app
 from absl import flags
 from absl import logging
+from six.moves import range
+from six.moves import urllib
+from six.moves import zip
 import tensorflow.compat.v1 as tf
 
 from mediapipe.util.sequence import media_sequence as ms
@@ -96,15 +98,15 @@ FILEPATTERN = "kinetics_700_%s_25fps_rgb_flow"
 SPLITS = {
     "train": {
         "shards": 1000,
-        "examples": 540247
+        "examples": 538779
     },
     "validate": {
         "shards": 100,
-        "examples": 34610
+        "examples": 34499
     },
     "test": {
         "shards": 100,
-        "examples": 69103
+        "examples": 68847
     },
     "custom": {
         "csv": None,  # Add a CSV for your own data here.
@@ -198,7 +200,7 @@ class Kinetics(object):
       return output_dict
 
     if split not in SPLITS:
-      raise ValueError("Split %s not in %s" % split, str(SPLITS.keys()))
+      raise ValueError("Split %s not in %s" % split, str(list(SPLITS.keys())))
     all_shards = tf.io.gfile.glob(
         os.path.join(self.path_to_data, FILEPATTERN % split + "-*-of-*"))
     random.shuffle(all_shards)
@@ -302,11 +304,12 @@ class Kinetics(object):
           continue
         # rename the row with a constitent set of names.
         if len(csv_row) == 5:
-          row = dict(zip(["label_name", "video", "start", "end", "split"],
-                         csv_row))
+          row = dict(
+              list(
+                  zip(["label_name", "video", "start", "end", "split"],
+                      csv_row)))
         else:
-          row = dict(zip(["video", "start", "end", "split"],
-                         csv_row))
+          row = dict(list(zip(["video", "start", "end", "split"], csv_row)))
         metadata = tf.train.SequenceExample()
         ms.set_example_id(bytes23(row["video"] + "_" + row["start"]),
                           metadata)
@@ -328,7 +331,7 @@ class Kinetics(object):
     if sys.version_info >= (3, 0):
       urlretrieve = urllib.request.urlretrieve
     else:
-      urlretrieve = urllib.urlretrieve
+      urlretrieve = urllib.request.urlretrieve
     logging.info("Creating data directory.")
     tf.io.gfile.makedirs(self.path_to_data)
     logging.info("Downloading annotations.")
@@ -404,7 +407,7 @@ class Kinetics(object):
         assert NUM_CLASSES == num_keys, (
             "Found %d labels for split: %s, should be %d" % (
                 num_keys, name, NUM_CLASSES))
-        label_map = dict(zip(classes, range(len(classes))))
+        label_map = dict(list(zip(classes, list(range(len(classes))))))
       if SPLITS[name]["examples"] > 0:
         assert SPLITS[name]["examples"] == num_examples, (
             "Found %d examples for split: %s, should be %d" % (

mediapipe/examples/desktop/youtube8m/README.md

@@ -30,6 +30,8 @@
     ```bash
     # cd to the root directory of the MediaPipe repo
     cd -
+
+    pip3 install tf_slim
     python -m mediapipe.examples.desktop.youtube8m.generate_vggish_frozen_graph
     ```
 
@@ -47,7 +49,7 @@
 5.  Run the MediaPipe binary to extract the features.
 
     ```bash
-    bazel build -c opt \
+    bazel build -c opt --linkopt=-s \
       --define MEDIAPIPE_DISABLE_GPU=1 --define no_aws_support=true \
       mediapipe/examples/desktop/youtube8m:extract_yt8m_features
 
@@ -87,7 +89,7 @@
 3.  Build and run the inference binary.
 
     ```bash
-    bazel build -c opt --define='MEDIAPIPE_DISABLE_GPU=1' \
+    bazel build -c opt --define='MEDIAPIPE_DISABLE_GPU=1' --linkopt=-s \
     mediapipe/examples/desktop/youtube8m:model_inference
 
     GLOG_logtostderr=1 bazel-bin/mediapipe/examples/desktop/youtube8m/model_inference \
@@ -113,13 +115,13 @@
 2.  Build the inference binary.
 
     ```bash
-    bazel build -c opt --define='MEDIAPIPE_DISABLE_GPU=1' \
+    bazel build -c opt --define='MEDIAPIPE_DISABLE_GPU=1' --linkopt=-s \
       mediapipe/examples/desktop/youtube8m:model_inference
     ```
 
 3.  Run the python web server.
 
-    Note: pip install absl-py
+    Note: pip3 install absl-py
 
     ```bash
     python mediapipe/examples/desktop/youtube8m/viewer/server.py --root `pwd`
@@ -142,7 +144,7 @@
 3.  Build and run the inference binary.
 
     ```bash
-    bazel build -c opt --define='MEDIAPIPE_DISABLE_GPU=1' \
+    bazel build -c opt --define='MEDIAPIPE_DISABLE_GPU=1' --linkopt=-s \
       mediapipe/examples/desktop/youtube8m:model_inference
 
     # segment_size is the number of seconds window of frames.

mediapipe/examples/desktop/youtube8m/generate_vggish_frozen_graph.py

@@ -25,7 +25,7 @@ import sys
 
 from absl import app
 import tensorflow.compat.v1 as tf
-from tensorflow.compat.v1.python.tools import freeze_graph
+from tensorflow.python.tools import freeze_graph
 
 BASE_DIR = '/tmp/mediapipe/'