diff --git a/map_nav_src/reverie/env.py b/map_nav_src/reverie/env.py
index 7234ebc..c89c175 100644
--- a/map_nav_src/reverie/env.py
+++ b/map_nav_src/reverie/env.py
@@ -8,12 +8,17 @@ import random
 import networkx as nx
 from collections import defaultdict
 import copy
+from glob import glob
+
 
 import MatterSim
 
 from utils.data import load_nav_graphs, new_simulator
 from utils.data import angle_feature, get_all_point_angle_feature
 
+with open('./node_region.json') as fp:
+    node_region = json.load(fp)
+
 
 class EnvBatch(object):
     ''' A simple wrapper for a batch of MatterSim environments,
@@ -360,6 +365,9 @@ class ReverieObjectNavBatch(object):
         path = sum(pred_path, [])
         assert gt_path[0] == path[0], 'Result trajectories should include the start position'
 
+        pred_stop_region = node_region[scan][path[-1]]
+        gt_stop_region = node_region[scan][gt_path[-1]]
+
         scores['action_steps'] = len(pred_path) - 1
         scores['trajectory_steps'] = len(path) - 1
         scores['trajectory_lengths'] = np.sum([shortest_distances[a][b] for a, b in zip(path[:-1], path[1:])])
@@ -369,15 +377,29 @@ class ReverieObjectNavBatch(object):
         goal_viewpoints = set(self.obj2vps['%s_%s'%(scan, str(gt_objid))])
         assert len(goal_viewpoints) > 0, '%s_%s'%(scan, str(gt_objid))
 
+        scores['found_success'] = float(pred_found == gt_found) 
+
         scores['success'] = float(path[-1] in goal_viewpoints)
+        scores['room_success'] = float(pred_stop_region == gt_stop_region)
+        scores['oracle_success'] = float(any(x in goal_viewpoints for x in path))
+        '''
         if scores['success'] == 1.0:
             scores['found_success'] = float(pred_found == gt_found) 
         else:
             scores['found_success'] = 0.0
+        '''
 
-        scores['oracle_success'] = float(any(x in goal_viewpoints for x in path))
+
+        
         scores['spl'] = scores['success'] * gt_lengths / max(scores['trajectory_lengths'], gt_lengths, 0.01)
-        scores['sspl'] = scores['spl'] * scores['found_success']
+        scores['sspl_1'] = scores['success'] * gt_lengths / max(scores['trajectory_lengths'], gt_lengths, 0.01) * scores['found_success']
+        scores['sspl_2'] = scores['room_success'] * gt_lengths / max(scores['trajectory_lengths'], gt_lengths, 0.01) * scores['found_success']
+        scores['sspl_3'] = scores['oracle_success'] * gt_lengths / max(scores['trajectory_lengths'], gt_lengths, 0.01) * scores['found_success']
+
+        scores['ss_1'] = scores['success'] * scores['found_success']
+        scores['ss_2'] = scores['room_success'] * scores['found_success']
+        scores['ss_3'] = scores['oracle_success'] * scores['found_success']
+        # scores['sspl'] = scores['spl'] * scores['found_success']
 
         scores['rgs'] = str(pred_objid) == str(gt_objid)
         scores['rgspl'] = scores['rgs'] * gt_lengths / max(scores['trajectory_lengths'], gt_lengths, 0.01)
@@ -419,8 +441,14 @@ class ReverieObjectNavBatch(object):
             'spl': np.mean(metrics['spl']) * 100,
             'rgs': np.mean(metrics['rgs']) * 100,
             'rgspl': np.mean(metrics['rgspl']) * 100,
-            'sspl': np.mean(metrics['sspl']) * 100,
+            'sspl_1': np.mean(metrics['sspl_1']) * 100,
+            'sspl_2': np.mean(metrics['sspl_2']) * 100,
+            'sspl_3': np.mean(metrics['sspl_3']) * 100,
+            'ss_1': np.mean(metrics['ss_1']) * 100,
+            'ss_2': np.mean(metrics['ss_2']) * 100,
+            'ss_3': np.mean(metrics['ss_3']) * 100,
             'found_sr': np.mean(metrics['found_success']) * 100,
+            'room_sr': np.mean(metrics['room_success']) * 100,
         }
         return avg_metrics, metrics