v1.3.6: added University-1652

README.rst

@@ -33,7 +33,7 @@ You can find some research projects that are built on top of Torchreid `here <ht
 
 What's new
 ------------
-- [Feb 2021] We support the new multi-view multi-source geo-localization dataset `University-1652 <https://dl.acm.org/doi/abs/10.1145/3394171.3413896>`_.
+- [Feb 2021] ``v1.3.6`` Added `University-1652 <https://dl.acm.org/doi/abs/10.1145/3394171.3413896>`_, a new dataset for multi-view multi-source geo-localization (credit to `Zhedong Zheng <https://github.com/layumi>`_).
 - [Feb 2021] ``v1.3.5``: Now the `cython code <https://github.com/KaiyangZhou/deep-person-reid/pull/412>`_ works on Windows (credit to `lablabla <https://github.com/lablabla>`_).
 - [Jan 2021] Our recent work, `MixStyle <https://openreview.net/forum?id=6xHJ37MVxxp>`_ (mixing instance-level feature statistics of samples of different domains for improving domain generalization), has been accepted to ICLR'21. The code has been released at https://github.com/KaiyangZhou/mixstyle-release where the person re-ID part is based on Torchreid.
 - [Jan 2021] A new evaluation metric called `mean Inverse Negative Penalty (mINP)` for person re-ID has been introduced in `Deep Learning for Person Re-identification: A Survey and Outlook (TPAMI 2021) <https://arxiv.org/abs/2001.04193>`_. Their code can be accessed at `<https://github.com/mangye16/ReID-Survey>`_.
@@ -232,7 +232,7 @@ Image-reid datasets
 - `PRID <https://pdfs.semanticscholar.org/4c1b/f0592be3e535faf256c95e27982db9b3d3d3.pdf>`_
 
 Geo-localization datasets
-^^^^^^^^^^^^^^^^^^^^^^^
+^^^^^^^^^^^^^^^^^^^^^^^^^^^
 - `University-1652 <https://dl.acm.org/doi/abs/10.1145/3394171.3413896>`_
 
 Video-reid datasets

torchreid/__init__.py

@@ -2,7 +2,7 @@ from __future__ import print_function, absolute_import
 
 from torchreid import data, optim, utils, engine, losses, models, metrics
 
-__version__ = '1.3.5'
+__version__ = '1.3.6'
 __author__ = 'Kaiyang Zhou'
 __homepage__ = 'https://kaiyangzhou.github.io/'
 __description__ = 'Deep learning person re-identification in PyTorch'

torchreid/data/datasets/__init__.py

@@ -2,7 +2,7 @@ from __future__ import print_function, absolute_import
 
 from .image import (
     GRID, PRID, CUHK01, CUHK02, CUHK03, MSMT17, VIPeR, SenseReID, Market1501,
-    DukeMTMCreID, iLIDS, University1652
+    DukeMTMCreID, University1652, iLIDS
 )
 from .video import PRID2011, Mars, DukeMTMCVidReID, iLIDSVID
 from .dataset import Dataset, ImageDataset, VideoDataset

torchreid/data/datasets/image/university1652.py

@@ -1,8 +1,8 @@
 from __future__ import division, print_function, absolute_import
+import os
 import re
 import glob
 import os.path as osp
-import os
 import gdown
 
 from ..dataset import ImageDataset
@@ -48,18 +48,23 @@ datamanager = torchreid.data.ImageDataManager(
     """
     dataset_dir = 'university1652'
     dataset_url = 'https://drive.google.com/uc?id=1iVnP4gjw-iHXa0KerZQ1IfIO0i1jADsR'
+
     def __init__(self, root='', **kwargs):
         self.root = osp.abspath(osp.expanduser(root))
         self.dataset_dir = osp.join(self.root, self.dataset_dir)
         print(self.dataset_dir)
         if not os.path.isdir(self.dataset_dir):
             os.mkdir(self.dataset_dir)
-            gdown.download(self.dataset_url, self.dataset_dir+'data.zip', quiet=False)
+            gdown.download(
+                self.dataset_url, self.dataset_dir + 'data.zip', quiet=False
+            )
             os.system('unzip %s' % (self.dataset_dir + 'data.zip'))
         self.train_dir = osp.join(
             self.dataset_dir, 'University-Release/train/'
         )
-        self.query_dir = osp.join(self.dataset_dir, 'University-Release/test/query_drone')
+        self.query_dir = osp.join(
+            self.dataset_dir, 'University-Release/test/query_drone'
+        )
         self.gallery_dir = osp.join(
             self.dataset_dir, 'University-Release/test/gallery_satellite'
         )
@@ -77,7 +82,10 @@ datamanager = torchreid.data.ImageDataManager(
         super(University1652, self).__init__(train, query, gallery, **kwargs)
 
     def process_dir(self, dir_path, relabel=False, train=False):
-        IMG_EXTENSIONS = ('.jpg', '.jpeg', '.png', '.ppm', '.bmp', '.pgm', '.tif', '.tiff', '.webp')
+        IMG_EXTENSIONS = (
+            '.jpg', '.jpeg', '.png', '.ppm', '.bmp', '.pgm', '.tif', '.tiff',
+            '.webp'
+        )
         if train:
             img_paths = glob.glob(osp.join(dir_path, '*/*/*'))
         else:

torchreid/utils/GPU-Re-Ranking/extension/adjacency_matrix/setup.py

@@ -16,22 +16,21 @@
     with limited time cost.
 """
 
-from setuptools import setup, Extension
-
+from setuptools import Extension, setup
 import torch
 import torch.nn as nn
 from torch.autograd import Function
-from torch.utils.cpp_extension import BuildExtension, CUDAExtension
-
+from torch.utils.cpp_extension import CUDAExtension, BuildExtension
 
 setup(
     name='build_adjacency_matrix',
     ext_modules=[
-        CUDAExtension('build_adjacency_matrix', [
+        CUDAExtension(
+            'build_adjacency_matrix', [
                 'build_adjacency_matrix.cpp',
                 'build_adjacency_matrix_kernel.cu',
-        ]),
+            ]
+        ),
     ],
-    cmdclass={
-        'build_ext':BuildExtension  
-    })
+    cmdclass={'build_ext': BuildExtension}
+)

torchreid/utils/GPU-Re-Ranking/extension/propagation/setup.py

@@ -16,22 +16,21 @@
     with limited time cost.
 """
 
-from setuptools import setup, Extension
-
+from setuptools import Extension, setup
 import torch
 import torch.nn as nn
 from torch.autograd import Function
-from torch.utils.cpp_extension import BuildExtension, CUDAExtension
-
+from torch.utils.cpp_extension import CUDAExtension, BuildExtension
 
 setup(
     name='gnn_propagate',
     ext_modules=[
-        CUDAExtension('gnn_propagate', [
+        CUDAExtension(
+            'gnn_propagate', [
                 'gnn_propagate.cpp',
                 'gnn_propagate_kernel.cu',
-        ]),
+            ]
+        ),
     ],
-    cmdclass={
-        'build_ext':BuildExtension  
-    })
+    cmdclass={'build_ext': BuildExtension}
+)

torchreid/utils/GPU-Re-Ranking/gnn_reranking.py

@@ -16,16 +16,14 @@
     with limited time cost.
 """
 
-import torch
 import numpy as np
+import torch
 
-import build_adjacency_matrix
 import gnn_propagate
-
+import build_adjacency_matrix
 from utils import *
 
 
-
 def gnn_reranking(X_q, X_g, k1, k2):
     query_num, gallery_num = X_q.shape[0], X_g.shape[0]
 
@@ -34,7 +32,9 @@ def gnn_reranking(X_q, X_g, k1, k2):
     del X_u, X_q, X_g
 
     # initial ranking list
-    S, initial_rank = original_score.topk(k=k1, dim=-1, largest=True, sorted=True)
+    S, initial_rank = original_score.topk(
+        k=k1, dim=-1, largest=True, sorted=True
+    )
 
     # stage 1
     A = build_adjacency_matrix.forward(initial_rank.float())
@@ -44,11 +44,13 @@ def gnn_reranking(X_q, X_g, k1, k2):
     if k2 != 1:
         for i in range(2):
             A = A + A.T
-            A = gnn_propagate.forward(A, initial_rank[:, :k2].contiguous().float(), S[:, :k2].contiguous().float())
+            A = gnn_propagate.forward(
+                A, initial_rank[:, :k2].contiguous().float(),
+                S[:, :k2].contiguous().float()
+            )
             A_norm = torch.norm(A, p=2, dim=1, keepdim=True)
             A = A.div(A_norm.expand_as(A))
 
-      
     cosine_similarity = torch.mm(A[:query_num, ], A[query_num:, ].t())
     del A, S
 

torchreid/utils/GPU-Re-Ranking/main.py

@@ -17,31 +17,38 @@
 """
 
 import os
-import torch
-import argparse
 import numpy as np
+import argparse
+import torch
 
 from utils import *
 from gnn_reranking import *
 
 parser = argparse.ArgumentParser(description='Reranking_is_GNN')
-parser.add_argument('--data_path', 
+parser.add_argument(
+    '--data_path',
     type=str,
     default='../xm_rerank_gpu_2/features/market_88_test.pkl',
-                    help='path to dataset')
-parser.add_argument('--k1', 
+    help='path to dataset'
+)
+parser.add_argument(
+    '--k1',
     type=int,
     default=26, # Market-1501
     # default=60,   # Veri-776
-                    help='parameter k1')
-parser.add_argument('--k2', 
+    help='parameter k1'
+)
+parser.add_argument(
+    '--k2',
     type=int,
     default=7, # Market-1501
     # default=10,   # Veri-776
-                    help='parameter k2')
+    help='parameter k2'
+)
 
 args = parser.parse_args()
 
+
 def main():
     data = load_pickle(args.data_path)
 
@@ -56,7 +63,10 @@ def main():
     gallery_feature = gallery_feature.cuda()
 
     indices = gnn_reranking(query_feature, gallery_feature, args.k1, args.k2)
-    evaluate_ranking_list(indices, query_label, query_cam, gallery_label, gallery_cam)
+    evaluate_ranking_list(
+        indices, query_label, query_cam, gallery_label, gallery_cam
+    )
+
 
 if __name__ == '__main__':
     main()

torchreid/utils/GPU-Re-Ranking/utils.py

@@ -16,9 +16,9 @@
     with limited time cost.
 """
 
-import pickle
-import numpy as np
 import os
+import numpy as np
+import pickle
 import torch
 
 
@@ -27,10 +27,12 @@ def load_pickle(pickle_path):
         data = pickle.load(f)
         return data
 
+
 def save_pickle(pickle_path, data):
     with open(pickle_path, 'wb') as f:
         pickle.dump(data, f, protocol=pickle.HIGHEST_PROTOCOL)
 
+
 def pairwise_squared_distance(x):
     '''
     x : (n_samples, n_points, dims)
@@ -39,17 +41,23 @@ def pairwise_squared_distance(x):
     x2s = (x * x).sum(-1, keepdim=True)
     return x2s + x2s.transpose(-1, -2) - 2 * x @ x.transpose(-1, -2)
 
+
 def pairwise_distance(x, y):
     m, n = x.size(0), y.size(0)
 
     x = x.view(m, -1)
     y = y.view(n, -1)
 
-    dist = torch.pow(x, 2).sum(dim=1, keepdim=True).expand(m, n) + torch.pow(y, 2).sum(dim=1, keepdim=True).expand(n,m).t()
+    dist = torch.pow(x, 2).sum(
+        dim=1, keepdim=True
+    ).expand(m, n) + torch.pow(y, 2).sum(
+        dim=1, keepdim=True
+    ).expand(n, m).t()
     dist.addmm_(1, -2, x, y.t())
 
     return dist
 
+
 def cosine_similarity(x, y):
     m, n = x.size(0), y.size(0)
 
@@ -61,12 +69,18 @@ def cosine_similarity(x, y):
 
     return score
 
-def evaluate_ranking_list(indices, query_label, query_cam, gallery_label, gallery_cam):   
+
+def evaluate_ranking_list(
+    indices, query_label, query_cam, gallery_label, gallery_cam
+):
     CMC = np.zeros((len(gallery_label)), dtype=np.int)
     ap = 0.0
 
     for i in range(len(query_label)):
-        ap_tmp, CMC_tmp = evaluate(indices[i],query_label[i], query_cam[i], gallery_label, gallery_cam)
+        ap_tmp, CMC_tmp = evaluate(
+            indices[i], query_label[i], query_cam[i], gallery_label,
+            gallery_cam
+        )
         if CMC_tmp[0] == -1:
             continue
         CMC = CMC + CMC_tmp
@@ -74,7 +88,11 @@ def evaluate_ranking_list(indices, query_label, query_cam, gallery_label, galler
 
     CMC = CMC.astype(np.float32)
     CMC = CMC / len(query_label) #average CMC
-    print('Rank@1:%f Rank@5:%f Rank@10:%f mAP:%f'%(CMC[0],CMC[4],CMC[9],ap/len(query_label)))
+    print(
+        'Rank@1:%f Rank@5:%f Rank@10:%f mAP:%f' %
+        (CMC[0], CMC[4], CMC[9], ap / len(query_label))
+    )
+
 
 def evaluate(index, ql, qc, gl, gc):
     query_index = np.argwhere(gl == ql)