Converted python 2 to python 3

lib/loaders/gt_mrcn_loader.py

@@ -247,15 +247,15 @@ def sample_neg_ids(self, ann_id, seq_per_ref, sample_ratio):
       elif len(dt_ann_ids) > 0:
         neg_ann_id = random.choice(dt_ann_ids)
       else:
-        neg_ann_id = random.choice(self.Anns.keys())
+        neg_ann_id = random.choice(list(self.Anns.keys()))
       neg_ann_ids += [neg_ann_id]
       # neg_ref_id for negative language representations: mainly from same-type "referred" objects
       if len(st_ref_ids) > 0 and np.random.uniform(0, 1, 1) < sample_ratio:
         neg_ref_id = random.choice(st_ref_ids)
       elif len(dt_ref_ids) > 0:
         neg_ref_id = random.choice(dt_ref_ids)
       else:
-        neg_ref_id = random.choice(self.Refs.keys())
+        neg_ref_id = random.choice(list(self.Refs.keys()))
       neg_sent_id = random.choice(self.Refs[neg_ref_id]['sent_ids'])
       neg_sent_ids += [neg_sent_id]
 
@@ -284,9 +284,29 @@ def compare(ann_id0, ann_id1):
       else:
         return 1
     image = self.Images[ref_ann['image_id']]        
-
+
+    def cmp_to_key(mycmp):
+        'Convert a cmp= function into a key= function'
+        class K(object):
+            def __init__(self, obj, *args):
+                self.obj = obj
+            def __lt__(self, other):
+                return mycmp(self.obj, other.obj) < 0
+            def __gt__(self, other):
+                return mycmp(self.obj, other.obj) > 0
+            def __eq__(self, other):
+                return mycmp(self.obj, other.obj) == 0
+            def __le__(self, other):
+                return mycmp(self.obj, other.obj) <= 0  
+            def __ge__(self, other):
+                return mycmp(self.obj, other.obj) >= 0
+            def __ne__(self, other):
+                return mycmp(self.obj, other.obj) != 0
+        return K
+
     ann_ids = list(image['ann_ids'])  # copy in case the raw list is changed
-    ann_ids = sorted(ann_ids, cmp=compare)
+    #ann_ids = sorted(ann_ids, cmp=compare)
+    ann_ids = sorted(ann_ids, key=cmp_to_key(compare))
 
     st_ref_ids, st_ann_ids, dt_ref_ids, dt_ann_ids = [], [], [], []
     for ann_id in ann_ids:

lib/mrcn/inference.py

@@ -144,6 +144,7 @@ def boxes_to_masks(self, img_path, boxes, labels):
     rles = []
     for m in masks:
       rle = COCOmask.encode(np.asfortranarray(m))
+      rle['counts'] = rle['counts'].decode('ascii')
       rles += [rle]
 
     return masks, rles

lib/mrcn/inference_no_imdb.py

@@ -259,7 +259,7 @@ def box_to_pool5_fc7(self, net_conv, im_info, ori_boxes):
 
     fc7 = self.net._head_to_tail(pool5)  # (n, 2048, 7, 7)
     pool5 = pool5.mean(3).mean(2)
-    fc7 = fc7.mean(3).mean(2)  # (n, 2048)
+    fc7 = fc7.mean(3, keepdim=True).mean(2, keepdim=True)  # (n, 2048)
     return pool5, fc7
 
   def box_to_fc7(self, net_conv, im_info, ori_boxes):

tools/eval_masks.py

@@ -23,6 +23,9 @@
 
 # compute IoU
 def computeIoU(pred_rle, gd_rle):
+  #pred_rle['counts'] = pred_rle['counts'].encode('ascii')
+  #gd_rle['counts'] = gd_rle['counts'].encode('ascii')
+
   pred_seg = maskUtils.decode(pred_rle) # (H, W)
   gd_seg = maskUtils.decode(gd_rle)     # (H, W)
   I = np.sum(np.logical_and(pred_seg, gd_seg))

tools/mattnet.py

@@ -306,7 +306,27 @@ def compare(det_id0, det_id1):
         return 1
 
     det_ids = list(Dets.keys())  # copy in case the raw list is changed
-    det_ids = sorted(det_ids, cmp=compare)
+
+    def cmp_to_key(mycmp):
+        #'Convert a cmp= function into a key= function'
+        class K(object):
+            def __init__(self, obj, *args):
+                self.obj = obj
+            def __lt__(self, other):
+                return mycmp(self.obj, other.obj) < 0
+            def __gt__(self, other):
+                return mycmp(self.obj, other.obj) > 0
+            def __eq__(self, other):
+                return mycmp(self.obj, other.obj) == 0
+            def __le__(self, other):
+                return mycmp(self.obj, other.obj) <= 0  
+            def __ge__(self, other):
+                return mycmp(self.obj, other.obj) >= 0
+            def __ne__(self, other):
+                return mycmp(self.obj, other.obj) != 0
+        return K
+
+    det_ids = sorted(det_ids, key=cmp_to_key(compare))
     st_det_ids, dt_det_ids = [], []
     for det_id in det_ids:
       if det_id != ref_det_id:

tools/prepro.py

@@ -68,7 +68,7 @@ def build_vocab(refer, params):
   vocab = good_words
 
   # add category words
-  category_names = refer.Cats.values() + ['__background__']
+  category_names = list(refer.Cats.values()) + ['__background__']
   for cat_name in category_names:
     for wd in cat_name.split():
       if wd not in word2count or word2count[wd] <= count_thr:
@@ -192,7 +192,7 @@ def build_att_vocab(refer, params, att_types=['r1', 'r2', 'r7']):
   sentToRef = refer.sentToRef
   ref_to_att_wds = {}
   forbidden = forbidden_noun + forbidden_att + forbidden_verb \
-              + refer.Cats.values() # we also forbid category name here
+              + list(refer.Cats.values()) # we also forbid category name here
   for sent in sents:
     sent_id = sent['sent_id']
     atts = sent['atts']

tools/run_detect.py

@@ -85,10 +85,10 @@ def main(args):
     scores, boxes = mrcn.predict(img_path)
 
     # get cls_to_dets, class_name -> [xyxysc] (n, 5)
-    cls_to_dets, num_dets = cls_to_detections(scores, boxes, imdb, args.nms_thresh, args.conf_thresh)
+    cls_to_dets, num_dets = cls_to_detections(scores, boxes, imdb, args.nms_thresh, float(args.conf_thresh))
 
     # make sure num_dets > 0 for this image, otherwise we lower down the conf_thresh
-    thresh = args.conf_thresh
+    thresh = float(args.conf_thresh)
     while num_dets == 0:
       thresh = thresh-0.1
       cls_to_dets, num_dets = cls_to_detections(scores, boxes, imdb, args.nms_thresh, thresh)

tools/run_detect_to_mask.py

@@ -59,6 +59,7 @@ def main(args):
   # load detections = [{det_id, box, image_id, category_id, category_name, score}]
   save_dir = osp.join('cache/detections', args.dataset+'_'+args.splitBy)
   detections = json.load(open(osp.join(save_dir, args.dets_file_name)))
+
   image_to_dets = {}
   for det in detections:
     image_id = det['image_id']
@@ -67,6 +68,7 @@ def main(args):
     image_to_dets[image_id] += [det]
 
   # run mask rcnn
+  #for i, image_id in enumerate(list(image_to_dets.keys())[:2]):
   for i, image_id in enumerate(image_to_dets.keys()):
     dets = image_to_dets[image_id]
 
@@ -89,6 +91,9 @@ def main(args):
 
   # save dets.json = [{det_id, box, image_id, score}]
   # to cache/detections/
+
+  #print(detections[0])
+
   save_path = osp.join(save_dir, args.dets_file_name[:-10] + '_masks.json')
   with open(save_path, 'w') as f:
     json.dump(detections, f)

tools/train.py

@@ -207,7 +207,8 @@ def main(args):
       infos['val_result_history'] = val_result_history
       infos['word_to_ix'] = loader.word_to_ix
       infos['att_to_ix'] = loader.att_to_ix
-      with open(osp.join(checkpoint_dir, opt['id']+'.json'), 'wb') as io:
+      #with open(osp.join(checkpoint_dir, opt['id']+'.json'), 'wb') as io:
+      with open(osp.join(checkpoint_dir, opt['id']+'.json'), 'w') as io:
         json.dump(infos, io)
 
     # update iter and epoch