[datasets] added the code to build the Multi-PIE PyTorch dataset

bob/learn/pytorch/datasets/__init__.py

+# gets sphinx autodoc done right - don't remove it
+__all__ = [_ for _ in dir() if not _.startswith('_')]
+

bob/learn/pytorch/datasets/multipie.py

+#!/usr/bin/env python
+# encoding: utf-8
+
+import os
+
+import torch
+import numpy
+
+from torch.utils.data import Dataset, DataLoader
+
+import bob.db.multipie
+import bob.io.base
+import bob.io.image
+
+
+class MultiPIEDataset(Dataset):
+  """MultiPIE dataset."""
+
+  def __init__(self, root_dir, world=False, transform=None):
+    """
+    """
+    self.root_dir = root_dir
+    self.transform = transform
+    self.world = world
+  
+    camera_to_pose = {'11_0': 'l90',
+                      '12_0': 'l75',
+                      '09_0': 'l60',
+                      '08_0': 'l45',
+                      '13_0': 'l30',
+                      '14_0': 'l15',
+                      '05_1': '0',
+                      '05_0': 'r15',
+                      '04_1': 'r30',
+                      '19_0': 'r45',
+                      '20_0': 'r60',
+                      '01_0': 'r75',
+                      '24_0': 'r90'}
+
+    camera_to_label = {'11_0': '0',
+                       '12_0': '1',
+                       '09_0': '2',
+                       '08_0': '3',
+                       '13_0': '4',
+                       '14_0': '5',
+                       '05_1': '6',
+                       '05_0': '7',
+                       '04_1': '8',
+                       '19_0': '9',
+                       '20_0': '10',
+                       '01_0': '11',
+                       '24_0': '12'}
+
+    # get all the needed file, the pose labels, and the id labels
+    self.data_files = []
+    self.pose_labels = []
+    id_labels = []
+
+    db = bob.db.multipie.Database()
+    
+    if world: 
+      c_set = db.clients(groups='world')
+    else:
+      c_set = db.clients()
+
+    ids = [client.id for client in c_set]
+    ids = sorted(ids, key=int)  
+
+    # filename and pose label are dependent on the camera
+    for camera in sorted(db.camera_names()):
+    
+      if world:
+        objs = db.objects(cameras=camera, groups='world')
+      else:
+        objs = db.objects(cameras=camera)
+   
+      # skip "high" cameras
+      if (camera == '19_1') or (camera == '08_1'):
+        continue
+
+      for obj in objs:
+        temp = os.path.split(obj.path)
+        identity = int(temp[0].split('/')[2])
+        id_labels.append(identity)
+        cropped_filename = os.path.join(root_dir, camera_to_pose[camera], temp[1])
+        cropped_filename += '.png'
+        self.data_files.append(cropped_filename)
+        self.pose_labels.append(camera_to_label[camera])
+      
+    self.id_labels = self.map_labels(id_labels)
+    print max(self.id_labels)
+      
+    
+  def __len__(self):
+    return len(self.data_files)
+ 
+
+  def __getitem__(self, idx):
+      image = bob.io.base.load(self.data_files[idx])
+      identity = self.id_labels[idx]
+      pose = self.pose_labels[idx]
+      sample = {'image': image, 'id': identity, 'pose': pose}
+
+      if self.transform:
+        sample = self.transform(sample)
+
+      return sample
+
+  def map_labels(self, raw_labels):
+    """
+    Map the clients to 0 to 1
+    """
+    
+    possible_labels = list(set(raw_labels))
+
+    labels = numpy.array(raw_labels)
+    
+    for i in range(len(possible_labels)):
+      l = possible_labels[i]
+      labels[numpy.where(labels==l)[0]] = i
+    
+    return labels
+