Add COCO val2014 (set) minus minival2014

data/README.md

@@ -48,6 +48,11 @@ Since you'll likely be experimenting with multiple installs of Fast/er R-CNN in
 parallel, you'll probably want to keep all of this data in a shared place and
 use symlinks. On my system I create the following symlinks inside `data`:
 
+Annotations for the 5k image 'minival' subset of COCO val2014 that I like to use
+can be found at http://www.cs.berkeley.edu/~rbg/faster-rcnn-data/instances_minival2014.json.zip.
+Annotations for COCO val2014 (set) minus minival (~35k images) can be found at
+http://www.cs.berkeley.edu/~rbg/faster-rcnn-data/instances_valminusminival2014.json.zip.
+
 ```
 # data/cache holds various outputs created by the datasets package
 ln -s /data/fast_rcnn_shared/cache

lib/datasets/coco.py

@@ -70,7 +70,10 @@ def __init__(self, image_set, year):
         # Some image sets are "views" (i.e. subsets) into others.
         # For example, minival2014 is a random 5000 image subset of val2014.
         # This mapping tells us where the view's images and proposals come from.
-        self._view_map = {'minival2014' : 'val2014'}
+        self._view_map = {
+            'minival2014' : 'val2014',          # 5k val2014 subset
+            'valminusminival2014' : 'val2014',  # val2014 \setminus minival2014
+        }
         coco_name = image_set + year  # e.g., "val2014"
         self._data_name = (self._view_map[coco_name]
                            if self._view_map.has_key(coco_name)
@@ -228,15 +231,25 @@ def _load_coco_annotation(self, index):
         handled by marking their overlaps (with all categories) to -1. This
         overlap value means that crowd "instances" are excluded from training.
         """
-        annIds = self._COCO.getAnnIds(imgIds=index, iscrowd=None)
-        objs = self._COCO.loadAnns(annIds)
-        objs = [obj for obj in objs if obj['area'] > 0]
-        num_objs = len(objs)
-
         im_ann = self._COCO.loadImgs(index)[0]
         width = im_ann['width']
         height = im_ann['height']
 
+        annIds = self._COCO.getAnnIds(imgIds=index, iscrowd=None)
+        objs = self._COCO.loadAnns(annIds)
+        # Sanitize bboxes -- some are invalid
+        valid_objs = []
+        for obj in objs:
+            x1 = np.max((0, obj['bbox'][0]))
+            y1 = np.max((0, obj['bbox'][1]))
+            x2 = np.min((width - 1, x1 + np.max((0, obj['bbox'][2] - 1))))
+            y2 = np.min((height - 1, y1 + np.max((0, obj['bbox'][3] - 1))))
+            if obj['area'] > 0 and x2 >= x1 and y2 >= y1:
+                obj['clean_bbox'] = [x1, y1, x2, y2]
+                valid_objs.append(obj)
+        objs = valid_objs
+        num_objs = len(objs)
+
         boxes = np.zeros((num_objs, 4), dtype=np.uint16)
         gt_classes = np.zeros((num_objs), dtype=np.int32)
         overlaps = np.zeros((num_objs, self.num_classes), dtype=np.float32)
@@ -249,12 +262,8 @@ def _load_coco_annotation(self, index):
                                          for cls in self._classes[1:]])
 
         for ix, obj in enumerate(objs):
-            x1 = obj['bbox'][0]
-            y1 = obj['bbox'][1]
-            x2 = np.min((width - 1, x1 + np.max((0, obj['bbox'][2] - 1))))
-            y2 = np.min((height - 1, y1 + np.max((0, obj['bbox'][3] - 1))))
             cls = coco_cat_id_to_class_ind[obj['category_id']]
-            boxes[ix, :] = [x1, y1, x2, y2]
+            boxes[ix, :] = obj['clean_bbox']
             gt_classes[ix] = cls
             seg_areas[ix] = obj['area']
             if obj['iscrowd']:

lib/datasets/factory.py

@@ -21,7 +21,7 @@
 
 # Set up coco_2014_<split>
 for year in ['2014']:
-    for split in ['train', 'val', 'minival']:
+    for split in ['train', 'val', 'minival', 'valminusminival']:
         name = 'coco_{}_{}'.format(year, split)
         __sets[name] = (lambda split=split, year=year: coco(split, year))