Using scene_id instead of scan_id

README.md

@@ -44,14 +44,14 @@ We used our method to generate 3D reconstructions in two different setups:
 For training IDR run:
 ```
 cd ./code
-python training/exp_runner.py --conf ./confs/dtu_fixed_cameras.conf --scan_id SCAN_ID
+python training/exp_runner.py --conf ./confs/dtu_fixed_cameras.conf --scene_id scene_id
 ```
-where SCAN_ID is the folder of the DTU scene to reconstruct, ie 'scan64'
+where scene_id is the folder of the DTU scene to reconstruct, ie 'scan64'
 
 Then, to produce the meshed surface, run:
 ```
 cd ./code
-python evaluation/eval.py  --conf ./confs/dtu_fixed_cameras.conf --scan_id SCAN_ID --checkpoint CHECKPOINT [--eval_rendering]
+python evaluation/eval.py  --conf ./confs/dtu_fixed_cameras.conf --scene_id scene_id --checkpoint CHECKPOINT [--eval_rendering]
 ```
 where CHECKPOINT is the epoch you wish to evaluate or 'latest' if you wish to take the most recent epoch.
 Turning on `--eval_rendering` will further produce and evaluate PSNR of train image reconstructions.
@@ -61,13 +61,13 @@ Turning on `--eval_rendering` will further produce and evaluate PSNR of train im
 For training IDR with cameras optimization run:
 ```
 cd ./code
-python training/exp_runner.py --train_cameras --conf ./confs/dtu_trained_cameras.conf --scan_id SCAN_ID
+python training/exp_runner.py --train_cameras --conf ./confs/dtu_trained_cameras.conf --scene_id scene_id
 ```
 
 Then, to evaluate cameras accuracy and to produce the meshed surface, run:
 ```
 cd ./code
-python evaluation/eval.py  --eval_cameras --conf ./confs/dtu_trained_cameras.conf --scan_id SCAN_ID --checkpoint CHECKPOINT [--eval_rendering]
+python evaluation/eval.py  --eval_cameras --conf ./confs/dtu_trained_cameras.conf --scene_id scene_id --checkpoint CHECKPOINT [--eval_rendering]
 ```
 
 
@@ -76,11 +76,11 @@ python evaluation/eval.py  --eval_cameras --conf ./confs/dtu_trained_cameras.con
 We have uploaded IDR trained models, and you can run the evaluation using:
 ```
 cd ./code
-python evaluation/eval.py --exps_folder trained_models --conf ./confs/dtu_fixed_cameras.conf --scan_id SCAN_ID  --checkpoint 2000 [--eval_rendering]
+python evaluation/eval.py --exps_folder trained_models --conf ./confs/dtu_fixed_cameras.conf --scene_id scene_id  --checkpoint 2000 [--eval_rendering]
 ```
 Or, for trained cameras:
 ```
-python evaluation/eval.py --exps_folder trained_models --conf ./confs/dtu_trained_cameras.conf --scan_id SCAN_ID --checkpoint 2000 --eval_cameras [--eval_rendering]
+python evaluation/eval.py --exps_folder trained_models --conf ./confs/dtu_trained_cameras.conf --scene_id scene_id --checkpoint 2000 --eval_cameras [--eval_rendering]
 ```
 
 

code/confs/ablation/dtu_fixed_noisy_cameras.conf

@@ -24,7 +24,7 @@ loss{
 dataset{
     data_dir = DTU
     img_res = [1200, 1600]
-    scan_id = scan114
+    scene_id = scan114
     cam_file = cameras_linear_init.npz
 }
 model{

code/confs/ablation/dtu_no_feature_vector.conf

@@ -24,7 +24,7 @@ loss{
 dataset{
     data_dir = DTU
     img_res = [1200, 1600]
-    scan_id = scan114
+    scene_id = scan114
 }
 model{
     feature_vector_size = 0

code/confs/ablation/dtu_no_normal.conf

@@ -24,7 +24,7 @@ loss{
 dataset{
     data_dir = DTU
     img_res = [1200, 1600]
-    scan_id = scan114
+    scene_id = scan114
 }
 model{
     feature_vector_size = 256

code/confs/ablation/dtu_no_view_dir.conf

@@ -24,7 +24,7 @@ loss{
 dataset{
     data_dir = DTU
     img_res = [1200, 1600]
-    scan_id = scan114
+    scene_id = scan114
 }
 model{
     feature_vector_size = 256

code/confs/crx_fixed_cameras.conf

@@ -24,7 +24,7 @@ loss{
 dataset{
     data_dir = DTU
     img_res = [600, 800]
-    scan_id = scan65
+    scene_id = scan65
 }
 model{
     feature_vector_size = 256

code/confs/crx_fixed_cameras_disentangled.conf

@@ -24,7 +24,7 @@ loss{
 dataset{
     data_dir = CRX
     img_res = [600, 804]
-    scan_id = scan65
+    scene_id = scan65
 }
 model{
     geometry_network

code/confs/crx_trained_cameras.conf

@@ -25,7 +25,7 @@ loss{
 dataset{
     data_dir = CRX
     img_res = [600, 804]
-    scan_id = scan65
+    scene_id = scan65
 }
 model{
     feature_vector_size = 256

code/confs/crx_trained_cameras_disentangled.conf

@@ -25,7 +25,7 @@ loss{
 dataset{
     data_dir = CRX
     img_res = [600, 804]
-    scan_id = scan65
+    scene_id = scan65
 }
 model{
     geometry_network

code/confs/crx_trained_cameras_prior.conf

@@ -25,7 +25,7 @@ loss{
 dataset{
     data_dir = CRX
     img_res = [600, 804]
-    scan_id = scan65
+    scene_id = scan65
 }
 model{
     geometry_network

code/confs/crx_trained_cameras_prior_latent.conf

@@ -25,7 +25,7 @@ loss{
 dataset{
     data_dir = CRX
     img_res = [600, 804]
-    scan_id = scan65
+    scene_id = scan65
 }
 model{
     geometry_network

code/confs/dtu_fixed_cameras.conf

@@ -24,7 +24,7 @@ loss{
 dataset{
     data_dir = DTU
     img_res = [1200, 1600]
-    scan_id = scan65
+    scene_id = scan65
 }
 model{
     feature_vector_size = 256

code/confs/dtu_trained_cameras.conf

@@ -25,7 +25,7 @@ loss{
 dataset{
     data_dir = DTU
     img_res = [1200, 1600]
-    scan_id = scan65
+    scene_id = scan65
 }
 model{
     feature_vector_size = 256

code/confs/dummy_fixed_cameras.conf

@@ -24,7 +24,7 @@ loss{
 dataset{
     data_dir = dummy
     img_res = [64, 64]
-    scan_id = scan0
+    scene_id = scan0
 }
 model{
     geometry_network

code/datasets/scene_dataset.py

@@ -12,11 +12,11 @@ def __init__(self,
                  train_cameras,
                  data_dir,
                  img_res,
-                 scan_id=0,
+                 scene_id='scan0',
                  cam_file=None
                  ):
 
-        self.instance_dir = os.path.join('../data', data_dir, '{0}'.format(scan_id))
+        self.instance_dir = os.path.join('../data', data_dir, '{0}'.format(scene_id))
 
         self.total_pixels = img_res[0] * img_res[1]
         self.img_res = img_res

code/evaluation/eval.py

@@ -24,9 +24,9 @@ def evaluate(**kwargs):
     eval_rendering = kwargs['eval_rendering']
 
     expname = conf.get_string('train.expname') + kwargs['expname']
-    scan_id = kwargs['scan_id'] if kwargs['scan_id'] else conf.get_string('dataset.scan_id', default=None)
-    if scan_id:
-        expname = expname + '_{0}'.format(scan_id)
+    scene_id = kwargs['scene_id'] if kwargs['scene_id'] else conf.get_string('dataset.scene_id', default=None)
+    if scene_id:
+        expname = expname + '_{0}'.format(scene_id)
 
     if kwargs['timestamp'] == 'latest':
         if os.path.exists(os.path.join('../', kwargs['exps_folder_name'], expname)):
@@ -52,8 +52,8 @@ def evaluate(**kwargs):
         model.cuda()
 
     dataset_conf = conf.get_config('dataset')
-    if kwargs['scan_id']:
-        dataset_conf['scan_id'] = kwargs['scan_id']
+    if kwargs['scene_id']:
+        dataset_conf['scene_id'] = kwargs['scene_id']
     eval_dataset = utils.get_class(conf.get_string('train.dataset_class'))(eval_cameras, **dataset_conf)
 
     # settings for camera optimization
@@ -173,7 +173,7 @@ def evaluate(**kwargs):
             psnrs.append(psnr)
 
         psnrs = np.array(psnrs).astype(np.float64)
-        print("RENDERING EVALUATION {2}: psnr mean = {0} ; psnr std = {1}".format("%.2f" % psnrs.mean(), "%.2f" % psnrs.std(), scan_id))
+        print("RENDERING EVALUATION {2}: psnr mean = {0} ; psnr std = {1}".format("%.2f" % psnrs.mean(), "%.2f" % psnrs.std(), scene_id))
 
 
 def get_cameras_accuracy(pred_Rs, gt_Rs, pred_ts, gt_ts,):
@@ -245,7 +245,7 @@ def calculate_psnr(img1, img2, mask):
     parser.add_argument('--gpu', type=str, default='auto', help='GPU to use [default: GPU auto]')
     parser.add_argument('--timestamp', default='latest', type=str, help='The experiemnt timestamp to test.')
     parser.add_argument('--checkpoint', default='latest',type=str,help='The trained model checkpoint to test')
-    parser.add_argument('--scan_id', type=str, help='If set, taken to be the scan id.')
+    parser.add_argument('--scene_id', type=str, help='If set, taken to be the scene id.')
     parser.add_argument('--resolution', default=512, type=int, help='Grid resolution for marching cube')
     parser.add_argument('--is_uniform_grid', default=False, action="store_true", help='If set, evaluate marching cube with uniform grid.')
     parser.add_argument('--eval_cameras', default=False, action="store_true", help='If set, evaluate camera accuracy of trained cameras.')
@@ -268,7 +268,7 @@ def calculate_psnr(img1, img2, mask):
              evals_folder_name='evals',
              timestamp=opt.timestamp,
              checkpoint=opt.checkpoint,
-             scan_id=opt.scan_id,
+             scene_id=opt.scene_id,
              resolution=opt.resolution,
              eval_cameras=opt.eval_cameras,
              eval_rendering=opt.eval_rendering

code/evaluation/eval_disentanglement.py

@@ -53,7 +53,7 @@ def evaluate(**kwargs):
     model.rendering_network = model_fake.rendering_network
 
     dataset_conf = conf.get_config('dataset')
-    dataset_conf['scan_id'] = geometry_id
+    dataset_conf['scene_id'] = geometry_id
     eval_dataset = utils.get_class(conf.get_string('train.dataset_class'))(False, **dataset_conf)
 
     eval_dataloader = torch.utils.data.DataLoader(eval_dataset,
@@ -138,8 +138,8 @@ def evaluate(**kwargs):
     parser = argparse.ArgumentParser()
     parser.add_argument('--conf', type=str, default='./confs/dtu_fixed_cameras.conf')
     parser.add_argument('--gpu', type=str, default='auto', help='GPU to use [default: GPU auto]')
-    parser.add_argument('--geometry_id', type=int, default=65, help='The scan id of the learned geometry.')
-    parser.add_argument('--appearance_id', type=int, default=110, help='The scan id of the learned appearance.')
+    parser.add_argument('--geometry_id', type=int, default='scan65', help='The scene id of the learned geometry.')
+    parser.add_argument('--appearance_id', type=int, default='scan110', help='The scene id of the learned appearance.')
 
     opt = parser.parse_args()
 

code/training/exp_runner.py

@@ -17,7 +17,7 @@
     parser.add_argument('--timestamp', default='latest', type=str, help='The timestamp of the run to be used in case of continuing from a previous run.')
     parser.add_argument('--checkpoint', default='latest',type=str,help='The checkpoint epoch number of the run to be used in case of continuing from a previous run.')
     parser.add_argument('--train_cameras', default=False, action="store_true", help='If set, optimizing also camera location.')
-    parser.add_argument('--scan_id', type=str, help='If set, taken to be the scan id.')
+    parser.add_argument('--scene_id', type=str, help='If set, taken to be the scene id.')
 
     opt = parser.parse_args()
 
@@ -36,7 +36,7 @@
                                  is_continue=opt.is_continue,
                                  timestamp=opt.timestamp,
                                  checkpoint=opt.checkpoint,
-                                 scan_id=opt.scan_id,
+                                 scene_id=opt.scene_id,
                                  train_cameras=opt.train_cameras
                                  )
 

code/training/idr_train.py

@@ -21,9 +21,9 @@ def __init__(self,**kwargs):
         self.train_cameras = kwargs['train_cameras']
 
         self.expname = self.conf.get_string('train.expname') + kwargs['expname']
-        scan_id = kwargs['scan_id'] if kwargs['scan_id'] else self.conf.get_string('dataset.scan_id', default=None)
-        if scan_id:
-            self.expname = self.expname + '_{0}'.format(scan_id)
+        scene_id = kwargs['scene_id'] if kwargs['scene_id'] else self.conf.get_string('dataset.scene_id', default=None)
+        if scene_id:
+            self.expname = self.expname + '_{0}'.format(scene_id)
 
         if kwargs['is_continue'] and kwargs['timestamp'] == 'latest':
             if os.path.exists(os.path.join('../',kwargs['exps_folder_name'],self.expname)):
@@ -78,8 +78,8 @@ def __init__(self,**kwargs):
         print('Loading data ...')
 
         dataset_conf = self.conf.get_config('dataset')
-        if kwargs['scan_id']:
-            dataset_conf['scan_id'] = kwargs['scan_id']
+        if kwargs['scene_id']:
+            dataset_conf['scene_id'] = kwargs['scene_id']
 
         self.train_dataset = utils.get_class(self.conf.get_string('train.dataset_class'))(self.train_cameras,
                                                                                           **dataset_conf)