Filter proposals not present in roidb

Summary:
It is convenient to run with a set of precomputed proposals for a
larger set (e.g., coco_trainval2014) and then do experiments using annotations
for a smaller subset of the data.

It is possible to do this filtering only if a config flag is set, but it seems
harmless enough to do it always.

Reviewed By: rbgirshick

Differential Revision: D14106953

fbshipit-source-id: f726425ee5b6b22fcb711e3ec479b302ea09e837

detectron/datasets/json_dataset.py

@@ -254,6 +254,8 @@ class JsonDataset(object):
         proposals = load_object(proposal_file)
 
         id_field = 'indexes' if 'indexes' in proposals else 'ids'  # compat fix
+
+        _remove_proposals_not_in_roidb(proposals, roidb, id_field)
         _sort_proposals(proposals, id_field)
         box_list = []
         for i, entry in enumerate(roidb):
@@ -453,3 +455,11 @@ def _sort_proposals(proposals, id_field):
     fields_to_sort = ['boxes', id_field, 'scores']
     for k in fields_to_sort:
         proposals[k] = [proposals[k][i] for i in order]
+
+
+def _remove_proposals_not_in_roidb(proposals, roidb, id_field):
+    # fix proposals so they don't contain entries for images not in the roidb
+    roidb_ids = set({entry["id"] for entry in roidb})
+    keep = [i for i, id in enumerate(proposals[id_field]) if id in roidb_ids]
+    for f in ['boxes', id_field, 'scores']:
+        proposals[f] = [proposals[f][i] for i in keep]