nicolasdonati / DUO-FM

Implementation of CVPR2022 paper: "Deep orientation-aware functional maps: Tackling symmetry issues in Shape Matching".
23 stars 5 forks source link

about scape dataset #9

Closed WangZhe1026 closed 6 months ago

WangZhe1026 commented 6 months ago

Hello, I had tried to reproduce this code via SMAL dataset successfully. But when I want to train it by SCAPE, it failes. I'm not clear about to slove this problem, maybe you can tell me how to set the correct train and test dataset? Because it only have test part of the SCAPE dataset. I tried to divide the test dataset into half and let another part be the training part and then the error message said :

Traceback (most recent call last): File "train.py", line 138, in train_net(cfg) File "train.py", line 39, in train_net train_dataset = ScapeDataset(dataset_path, name=cfg["dataset"]["name"]+"-"+cfg["dataset"]["subset"], File "/home/ma3/wz/DUO-FM/scape_dataset.py", line 178, in init mesh_for_Q = qm.mesh(str(mesh_dirpath / f"{shape_name}{ext}"), File "/home/ma3/wz/DUO-FM/Tools/mesh.py", line 89, in init self.complex_spectral(k=complex_spectral, spectral_folder=spectral_folder) File "/home/ma3/wz/DUO-FM/Tools/mesh.py", line 206, in complex_spectral self.cl = self.complex_laplacian() File "/home/ma3/wz/DUO-FM/Tools/mesh.py", line 338, in complex_laplacian self.local_basis() File "/home/ma3/wz/DUO-FM/Tools/mesh.py", line 249, in local_basis self.halfedge() File "/home/ma3/wz/DUO-FM/Tools/mesh.py", line 118, in halfedge e2ue2e = np.array(ue2e)[emap] ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (19396,) + inhomogeneous part.

Maybe it will help, thank you if you can help me solve this problem.