update

app.py

@@ -20,11 +20,13 @@ with open('shape_names.txt') as f:
 model_gda = GDANET()
 model_gda = nn.DataParallel(model_gda)
 model_gda.load_state_dict(torch.load('./GDANet_WOLFMix.t7', map_location=torch.device('cpu')))
+# model_gda.load_state_dict(torch.load('/Users/renjiawei/Downloads/pretrained_models/GDANet_WOLFMix.t7', map_location=torch.device('cpu')))
 model_gda.eval()
 
 model_dgcnn = DGCNN()
 model_dgcnn = nn.DataParallel(model_dgcnn)
 model_dgcnn.load_state_dict(torch.load('./dgcnn.t7', map_location=torch.device('cpu')))
+# model_dgcnn.load_state_dict(torch.load('/Users/renjiawei/Downloads/pretrained_models/dgcnn.t7', map_location=torch.device('cpu')))
 model_dgcnn.eval()
 
 def pyplot_draw_point_cloud(points, corruption):
@@ -67,8 +69,10 @@ def load_dataset(corruption_idx, severity):
     corruption_type = corruptions[corruption_idx]
     if corruption_type == 'clean':
         f = h5py.File(osp.join('modelnet_c', corruption_type + '.h5'))
+        # f = h5py.File(osp.join('/Users/renjiawei/Downloads/modelnet_c', corruption_type + '.h5'))
     else:
         f = h5py.File(osp.join('modelnet_c', corruption_type + '_{}'.format(severity-1) + '.h5'))
+        # f = h5py.File(osp.join('/Users/renjiawei/Downloads/modelnet_c', corruption_type + '_{}'.format(severity - 1) + '.h5'))
     data = f['data'][:].astype('float32')
     label = f['label'][:].astype('int64')
     f.close()
@@ -91,6 +95,22 @@ def run(seed, corruption_idx, severity):
     output = 'visualization.png'
     return output, recognize_pcd(model_dgcnn, pcd), recognize_pcd(model_gda, pcd)
 
+
+description = """
+
+Welcome to the demo of PointCloud-C! PointCloud-C is the very first test-suite for point cloud robustness analysis under corruptions.
+
+In this demo, you may:
+
+- __Visualize__ various types of corrupted point clouds in [ModelNet-C](https://github.com/jiawei-ren/ModelNet-C).
+
+- __Compare__ our proposed techniques to the baseline in terms of prediction robustness.
+
+For more details, checkout our paper [Benchmarking and Analyzing Point Cloud Classification under Corruptions, __ICML 2022__](https://arxiv.org/abs/2202.03377) and our [project page](https://pointcloud-c.github.io/home.html)!
+
+"""
+
+
 if __name__ == '__main__':
     iface = gr.Interface(
         fn=run,
@@ -108,23 +128,13 @@ if __name__ == '__main__':
         ],
         live=False,
         allow_flagging='never',
-        title="ModelNet-C",
-        description="""
-        
-                    Welcome to the demo of ModelNet-C! In this demo, you may: 
-                    
-                    - **Visualize** various types of corrupted point clouds in ModelNet-C,
-                    
-                    - **Compare** our proposed techniques to the baseline in terms of prediction robustness.
-                    
-                    For more details, checkout more our paper [Benchmarking and Analyzing Point Cloud Classification under Corruptions, ICML 2022](https://arxiv.org/abs/2202.03377)!
-                    
-                    """,
+        title="PointCloud-C",
+        description=description,
         examples=[
             [0, 'Jitter', 5],
             [999, 'Drop Local', 5],
         ],
-        # css=".output-image, .image-preview {height: 500px !important}",
-        article="<p style='text-align: center'><a href='https://github.com/jiawei-ren/ModelNet-C' target='_blank'>ModelNet-C @ GitHub</a></p> "
+        css=".output-image, .image-preview {height: 100px !important}",
+        article="<p style='text-align: center'><a href='https://github.com/ldkong1205/PointCloud-C target='_blank'>PointNet-C @ GitHub</a></p> "
     )
     iface.launch()