model change

Author: joonson

README.md

@@ -36,9 +36,9 @@ python demo_syncnet.py --videofile data/example.avi --tmp_dir /path/to/temp/dire
 
 Check that this script returns:
 ```
-AV offset:      4 
-Min dist:       6.742
-Confidence:     10.447
+AV offset:      3 
+Min dist:       5.353
+Confidence:     10.021
 ```
 
 Full pipeline:

SyncNetInstance.py

@@ -55,8 +55,7 @@ def evaluate(self, opt, videofile):
             if ret == 0:
                 break
 
-            image_np = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
-            images.append(image_np)
+            images.append(image)
 
         im = numpy.stack(images,axis=3)
         im = numpy.expand_dims(im,axis=0)

SyncNetModel.py

@@ -42,23 +42,23 @@ def __init__(self, num_layers_in_fc_layers = 1024):
             nn.ReLU(inplace=True),
             nn.MaxPool2d(kernel_size=(3,3), stride=(2,2)),
 
-            nn.Conv2d(256, 2048, kernel_size=(5,4), padding=(0,0)),
-            nn.BatchNorm2d(2048),
+            nn.Conv2d(256, 512, kernel_size=(5,4), padding=(0,0)),
+            nn.BatchNorm2d(512),
             nn.ReLU(),
         );
 
         self.netfcaud = nn.Sequential(
-            nn.Linear(2048, 2048),
-            nn.BatchNorm1d(2048),
+            nn.Linear(512, 512),
+            nn.BatchNorm1d(512),
             nn.ReLU(),
-            nn.Linear(2048, num_layers_in_fc_layers),
+            nn.Linear(512, num_layers_in_fc_layers),
         );
 
         self.netfclip = nn.Sequential(
-            nn.Linear(2048, 2048),
-            nn.BatchNorm1d(2048),
+            nn.Linear(512, 512),
+            nn.BatchNorm1d(512),
             nn.ReLU(),
-            nn.Linear(2048, num_layers_in_fc_layers),
+            nn.Linear(512, num_layers_in_fc_layers),
         );
 
         self.netcnnlip = nn.Sequential(
@@ -85,8 +85,8 @@ def __init__(self, num_layers_in_fc_layers = 1024):
             nn.ReLU(inplace=True),
             nn.MaxPool3d(kernel_size=(1,3,3), stride=(1,2,2)),
 
-            nn.Conv3d(256, 2048, kernel_size=(1,6,6), padding=0),
-            nn.BatchNorm3d(2048),
+            nn.Conv3d(256, 512, kernel_size=(1,6,6), padding=0),
+            nn.BatchNorm3d(512),
             nn.ReLU(inplace=True),
         );
 

demo_feature.py

@@ -10,7 +10,7 @@
 
 parser = argparse.ArgumentParser(description = "SyncNet");
 
-parser.add_argument('--initial_model', type=str, default="data/syncnetl2.model", help='');
+parser.add_argument('--initial_model', type=str, default="data/syncnet_v2.model", help='');
 parser.add_argument('--batch_size', type=int, default='20', help='');
 parser.add_argument('--vshift', type=int, default='15', help='');
 parser.add_argument('--videofile', type=str, default="data/example.avi", help='');

demo_syncnet.py

@@ -10,7 +10,7 @@
 
 parser = argparse.ArgumentParser(description = "SyncNet");
 
-parser.add_argument('--initial_model', type=str, default="data/syncnetl2.model", help='');
+parser.add_argument('--initial_model', type=str, default="data/syncnet_v2.model", help='');
 parser.add_argument('--batch_size', type=int, default='20', help='');
 parser.add_argument('--vshift', type=int, default='15', help='');
 parser.add_argument('--videofile', type=str, default="data/example.avi", help='');

download_model.sh

@@ -1,7 +1,7 @@
 # SyncNet model
 
 mkdir data
-wget http://www.robots.ox.ac.uk/~vgg/software/lipsync/data/syncnetl2.model -O data/syncnetl2.model
+wget http://www.robots.ox.ac.uk/~vgg/software/lipsync/data/syncnet_v2.model -O data/syncnet_v2.model
 wget http://www.robots.ox.ac.uk/~vgg/software/lipsync/data/example.avi -O data/example.avi
 
 # For the pre-processing pipeline

run_syncnet.py

@@ -8,7 +8,7 @@
 # ==================== PARSE ARGUMENT ====================
 
 parser = argparse.ArgumentParser(description = "SyncNet");
-parser.add_argument('--initial_model', type=str, default="data/syncnetl2.model", help='');
+parser.add_argument('--initial_model', type=str, default="data/syncnet_v2.model", help='');
 parser.add_argument('--batch_size', type=int, default='20', help='');
 parser.add_argument('--vshift', type=int, default='15', help='');
 parser.add_argument('--data_dir', type=str, default='/mnt/hdd1/krdemo4', help='');