add compute_mean_std tool

tools/compute_mean_std.py

@@ -0,0 +1,60 @@
+import torchreid
+import argparse
+
+
+"""
+Compute channel-wise mean and standard deviation of a dataset.
+
+Usage:
+$ python compute_mean_std.py DATASET_ROOT DATASET_KEY
+
+- The first argument points to the root path where you put the datasets.
+- The second argument means the specific dataset key.
+
+For instance, your datasets are put under $DATA and you wanna
+compute the statistics of Market1501, do
+$ python compute_mean_std.py $DATA market1501
+"""
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('root', type=str)
+    parser.add_argument('sources', type=str)
+    args = parser.parse_args()
+
+    datamanager = torchreid.data.ImageDataManager(
+        root=args.root,
+        sources=args.sources,
+        targets=None,
+        height=256,
+        width=128,
+        batch_size_train=100,
+        batch_size_test=100,
+        transforms=None,
+        norm_mean=[0., 0., 0.],
+        norm_std=[1., 1., 1.],
+        train_sampler='SequentialSampler'
+    )
+    train_loader = datamanager.train_loader
+
+    print('Computing mean and std ...')
+    mean = 0.
+    std = 0.
+    n_samples = 0.
+    for data in train_loader:
+        data = data[0]
+        batch_size = data.size(0)
+        data = data.view(batch_size, data.size(1), -1)
+        mean += data.mean(2).sum(0)
+        std += data.std(2).sum(0)
+        n_samples += batch_size
+
+    mean /= n_samples
+    std /= n_samples
+    print('Mean: {}'.format(mean))
+    print('Std: {}'.format(std))
+
+
+if __name__ == '__main__':
+    main()

torchreid/data/datamanager.py

@@ -100,7 +100,7 @@ class ImageDataManager(DataManager):
         workers (int, optional): number of workers. Default is 4.
         num_instances (int, optional): number of instances per identity in a batch.
             Default is 4.
-        train_sampler (str, optional): sampler. Default is empty (``RandomSampler``).
+        train_sampler (str, optional): sampler. Default is RandomSampler.
         cuhk03_labeled (bool, optional): use cuhk03 labeled images.
             Default is False (defaul is to use detected images).
         cuhk03_classic_split (bool, optional): use the classic split in cuhk03.
@@ -148,7 +148,7 @@ class ImageDataManager(DataManager):
         batch_size_test=32,
         workers=4,
         num_instances=4,
-        train_sampler='',
+        train_sampler='RandomSampler',
         cuhk03_labeled=False,
         cuhk03_classic_split=False,
         market1501_500k=False
@@ -321,7 +321,7 @@ class VideoDataManager(DataManager):
         workers (int, optional): number of workers. Default is 4.
         num_instances (int, optional): number of instances per identity in a batch.
             Default is 4.
-        train_sampler (str, optional): sampler. Default is empty (``RandomSampler``).
+        train_sampler (str, optional): sampler. Default is RandomSampler.
         seq_len (int, optional): how many images to sample in a tracklet. Default is 15.
         sample_method (str, optional): how to sample images in a tracklet. Default is "evenly".
             Choices are ["evenly", "random", "all"]. "evenly" and "random" will sample ``seq_len``
@@ -372,7 +372,7 @@ class VideoDataManager(DataManager):
         batch_size_test=3,
         workers=4,
         num_instances=4,
-        train_sampler=None,
+        train_sampler='RandomSampler',
         seq_len=15,
         sample_method='evenly'
     ):

torchreid/data/sampler.py

@@ -95,7 +95,7 @@ def build_train_sampler(data_source, train_sampler, batch_size=32, num_instances
     elif train_sampler == 'SequentialSampler':
         sampler = SequentialSampler(data_source)
     
-    else:
+    elif train_sampler == 'RandomSampler':
         sampler = RandomSampler(data_source)
 
     return sampler