Training updated

lojzezust · Jul 23, 2022 · f3ef327 · f3ef327
1 parent fea97f6
commit f3ef327
Show file tree

Hide file tree

Showing 4 changed files with 3 additions and 10 deletions.
diff --git a/README.md b/README.md
@@ -7,7 +7,7 @@ Our work will be presented at the *IROS 2022* conference in Kyoto, Japan.
 
 <p align="center">
     <img src="figures/comparison.gif" alt="Comparison WaSR - WaSR-T">
-    Comparison between WaSR (single-frame) and WaSR-T (temporal context).
+    Comparison between WaSR (single-frame) and WaSR-T (temporal context) on hard examples.
 </p>
 
 

diff --git a/train.py b/train.py
@@ -85,8 +85,6 @@ def get_arguments(input_args=None):
                         help="Name of the model. Used to create model and log directories inside the output directory.")
     parser.add_argument("--pretrained-weights", type=str, default=None,
                         help="Path to the pretrained weights to be used.")
-    parser.add_argument("--architecture", type=str, choices=M.models, default=ARCHITECTURE,
-                        help="Which architecture to use.")
     parser.add_argument("--monitor-metric", type=str, default=MONITOR_VAR,
                         help="Validation metric to monitor for early stopping and best model saving.")
     parser.add_argument("--monitor-metric-mode", type=str, default=MONITOR_VAR_MODE, choices=['min', 'max'],

diff --git a/wasr_t/train.py b/wasr_t/train.py
@@ -8,7 +8,7 @@
 from .loss import focal_loss, water_obstacle_separation_loss
 from .metrics import PixelAccuracy, ClassIoU
 
-NUM_EPOCHS = 50
+NUM_EPOCHS = 100
 LEARNING_RATE = 1e-6
 MOMENTUM = 0.9
 WEIGHT_DECAY = 1e-6

diff --git a/wasr_t/wasr_t.py b/wasr_t/wasr_t.py
@@ -44,11 +44,9 @@ def wasr_temporal_resnet101(num_classes=3, pretrained=True, sequential=False):
 
 class WaSRT(nn.Module):
     """WaSR-T model"""
-    def __init__(self, backbone, decoder, backbone_grad_steps=3, imu=False, sequential=False):
+    def __init__(self, backbone, decoder, backbone_grad_steps=3, sequential=False):
         super(WaSRT, self).__init__()
 
-        self.imu = imu
-
         self.backbone = backbone
         self.decoder = decoder
         self.backbone_grad_steps = backbone_grad_steps
@@ -92,9 +90,6 @@ def forward_unrolled(self, x):
         for f in extract_feats:
             feats_hist[f] = torch.stack(feats_hist[f], 1)
 
-        if self.imu:
-            features['imu_mask'] = x['imu_mask']
-
         x = self.decoder(features, feats_hist)
 
         # Return segmentation map and aux feature map