“ Bug fix”

lib/datasets/pascal_voc.py

@@ -46,11 +46,7 @@ def __init__(self, image_set, year, devkit_path=None):
             else devkit_path
         self._data_path = os.path.join(self._devkit_path, 'VOC' + self._year)
         self._classes = ('__background__',  # always index 0
-                         'aeroplane', 'bicycle', 'bird', 'boat',
-                         'bottle', 'bus', 'car', 'cat', 'chair',
-                         'cow', 'diningtable', 'dog', 'horse',
-                         'motorbike', 'person', 'pottedplant',
-                         'sheep', 'sofa', 'train', 'tvmonitor')
+                        'person','car','van','truck','misc','dontcare','cyclist','tram','person_sitting')
         self._class_to_ind = dict(zip(self.classes, xrange(self.num_classes)))
         self._image_ext = '.jpg'
         self._image_index = self._load_image_set_index()

lib/model/faster_rcnn/faster_rcnn.py

@@ -19,7 +19,7 @@
 class _fasterRCNN(nn.Module):
     """ faster RCNN """
     def __init__(self, classes, class_agnostic):
-        super(_fasterRCNN, self).__init__()
+        super(_fasterRCNN, self).__init__() #继承Module的舒适化
         self.classes = classes
         self.n_classes = len(classes)
         self.class_agnostic = class_agnostic
@@ -33,7 +33,7 @@ def __init__(self, classes, class_agnostic):
         self.RCNN_roi_pool = _RoIPooling(cfg.POOLING_SIZE, cfg.POOLING_SIZE, 1.0/16.0)
         self.RCNN_roi_align = RoIAlignAvg(cfg.POOLING_SIZE, cfg.POOLING_SIZE, 1.0/16.0)
 
-        self.grid_size = cfg.POOLING_SIZE * 2 if cfg.CROP_RESIZE_WITH_MAX_POOL else cfg.POOLING_SIZE
+        self.grid_size = cfg.POOLING_SIZE * 2 if cfg.CROP_RESIZE_WITH_MAX_POOL else cfg.POOLING_SIZE #def larger(num1, num2): return num1 if num1 > num2 else num2, this variable may be declare in other place
         self.RCNN_roi_crop = _RoICrop()
 
     def forward(self, im_data, im_info, gt_boxes, num_boxes):
@@ -77,13 +77,13 @@ def forward(self, im_data, im_info, gt_boxes, num_boxes):
             pooled_feat = self.RCNN_roi_crop(base_feat, Variable(grid_yx).detach())
             if cfg.CROP_RESIZE_WITH_MAX_POOL:
                 pooled_feat = F.max_pool2d(pooled_feat, 2, 2)
-        elif cfg.POOLING_MODE == 'align':
+        elif cfg.POOLING_MODE == 'align': #this is the most common ROI Pooling method.
             pooled_feat = self.RCNN_roi_align(base_feat, rois.view(-1, 5))
         elif cfg.POOLING_MODE == 'pool':
             pooled_feat = self.RCNN_roi_pool(base_feat, rois.view(-1,5))
 
         # feed pooled features to top model
-        pooled_feat = self._head_to_tail(pooled_feat)
+        pooled_feat = self._head_to_tail(pooled_feat)   #problem1
 
         # compute bbox offset
         bbox_pred = self.RCNN_bbox_pred(pooled_feat)

lib/model/faster_rcnn/vgg16.py

@@ -35,7 +35,7 @@ def _init_modules(self):
     vgg.classifier = nn.Sequential(*list(vgg.classifier._modules.values())[:-1])
 
     # not using the last maxpool layer
-    self.RCNN_base = nn.Sequential(*list(vgg.features._modules.values())[:-1])
+    self.RCNN_base = nn.Sequential(*list(vgg.features._modules.values())[:-1])  #[:-1]是去掉最后个元素的意思
 
     # Fix the layers before conv3:
     for layer in range(10):
@@ -47,9 +47,10 @@ def _init_modules(self):
 
     # not using the last maxpool layer
     self.RCNN_cls_score = nn.Linear(4096, self.n_classes)
-
+        
     if self.class_agnostic:
-      self.RCNN_bbox_pred = nn.Linear(4096, 4)
+       #important!!!!!!!!!
+      self.RCNN_bbox_pred = nn.Linear(4096, 4) #LINEAR is full connection layer, 4096 is input , 4 is output，相当于最后一个全联接层1000个输出那一项去掉。
     else:
       self.RCNN_bbox_pred = nn.Linear(4096, 4 * self.n_classes)      
 

lib/roi_data_layer/roibatchLoader.py

@@ -50,8 +50,8 @@ def __init__(self, roidb, ratio_list, ratio_index, batch_size, num_classes, trai
         else:
             # for ratio cross 1, we make it to be 1.
             target_ratio = 1
-
-        self.ratio_list_batch[left_idx:(right_idx+1)] = target_ratio
+        temp = torch.ones(batch_size)*target_ratio
+        self.ratio_list_batch[left_idx:(right_idx+1)] = temp
 
 
   def __getitem__(self, index):