Patch 2

src/integrationtest/python/system_tests.py

@@ -1,6 +1,6 @@
 import os
 import tempfile
-import sh
+import subprocess
 import json
 import atexit
 import shutil
@@ -59,14 +59,14 @@ def check_generated_files():
         source_frames.write(b'\x00' * 30000)
         source_frames.write(b'\xff' * 30000)
 
-sh.ffmpeg('-t', '10',
+subprocess.call(['ffmpeg','-t', '10',
           '-s', '100x100',
           '-f', 'rawvideo',
           '-pix_fmt', 'rgb24',
           '-r', '8',
           '-i', './source',
-          'source.mp4',
-          _cwd=temp_dir,
+          'source.mp4']
+          cwd=temp_dir,
           )
 
 json_content = []
@@ -86,11 +86,11 @@ def check_generated_files():
 # step 2: run the gulping
 
 # PATH=src/main/scripts:$PATH PYTHONPATH=src/main/python
-command = sh.gulp_20bn_json_videos(
+command = subprocess.call(['gulp_20bn_json_videos',
     '--videos_per_chunk',  '10',
     os.path.join(temp_dir, 'videos.json'),
     temp_dir,
-    output_dir,
+    output_dir]
 )
 
 # step 3: sanity check the output
@@ -123,11 +123,11 @@ def check_generated_files():
 
 # step 6: extend the existing gulps
 
-command = sh.gulp_20bn_json_videos(
+command = subprocess.call(['gulp_20bn_json_videos',
     '--videos_per_chunk',  '10',
     os.path.join(temp_dir, 'videos_extend.json'),
     temp_dir,
-    output_dir,
+    output_dir]
 )
 
 # step 7: sanity check the extended output

src/main/python/gulpio/dataset.py

@@ -17,16 +17,16 @@ def __init__(self, data_path, num_frames, step_size,
 
             Args:
                 data_path (str): path to GulpIO dataset folder
-                label_path (str): path to GulpIO label dictionary matching
-            label ids to label names
                 num_frames (int): number of frames to be fetched.
                 step_size (int): number of frames skippid while picking
             sequence of frames from each video.
-                is_va (bool): sets the necessary augmention procedure.
+                is_val (bool): sets the necessary augmention procedure.
                 transform (object): set of augmentation steps defined by
             Compose(). Default is None.
-                target_transform (func): performs preprocessing on labels if
-            defined. Default is None.
+                target_transform (func):  a transformation function applied to each
+            single target, where target is the id assigned to a label. The
+            mapping from label to id is provided in the `label_idx` member-
+            variable. Default is None.
                 stack (bool): stack frames into a numpy.array. Default is True.
                 random_offset (bool): random offsetting to pick frames, if
             number of frames are more than what is necessary.
@@ -88,6 +88,8 @@ def __getitem__(self, index):
         # augmentation
         if self.transform_video:
             frames = self.transform_video(frames)
+        if self.target_transform:
+            target_idx = self.target_transform(target_idx)
         # format data to torch tensor
         if self.stack:
             frames = np.stack(frames)

src/unittest/python/loader_tests.py

@@ -84,6 +84,13 @@ def test_dataset(self):
                                    False, stack=False)
         self.iterate(loader)
 
+    def test_target_transform(self):
+        self.create_chunk()
+        target_label = 7
+        dataset = GulpVideoDataset(self.temp_dir, 2, 2,
+                                   False, target_transform=lambda y: target_label)
+        assert dataset[0][1] == target_label
+
 
 class TestGulpImageDataset(unittest.TestCase):