czq142857 / IM-NET-pytorch

PyTorch 1.2 implementation of IM-NET.
Other
196 stars 27 forks source link

About test #15

Closed Kammagod closed 3 years ago

Kammagod commented 3 years ago

hi, I want to input only one image without his voxel to reconstruction, is that feasible?

czq142857 commented 3 years ago

Yes. Please take a look at function test_image at modelSVR.py.

Kammagod commented 3 years ago

Yes. Please take a look at function _testimage at modelSVR.py.

Thank you for your help,but in Class IM_SVR,data_hdf5 file is generated using voxels and images. Can I use only images to generate this file? Please give me some suggestions

czq142857 commented 3 years ago

You can simply remove the data loading code, since the training data is not used in test_image.

Kammagod commented 3 years ago

Thank you! but in test, I should use self.dataset_load = self.dataset_name + '_test'and data_hdf5_name = self.data_dir+self.dataset_load+'.hdf5',i mean whether all_vox256_img_test.hdf5 file could be made just by images without their voxels?

czq142857 commented 3 years ago

Yes, you can generate an hdf5 file with only 'pixels'. You can modify the data preparation code to achieve that. But why do you need an hdf5 file if you can directly input images using test_image?

Kammagod commented 3 years ago

Th

Yes, you can generate an hdf5 file with only 'pixels'. You can modify the data preparation code to achieve that. But why do you need an hdf5 file if you can directly input images using _testimage?

There is no 'test_image' in class IM_SVR... In the code downloaded, there is only 'test_mesh' and 'test_mesh_point' method in class IM_SCR, which one is 'test_image'?

czq142857 commented 3 years ago

Please clone this repo again. I added this function this morning after seeing this issue.

Kammagod commented 3 years ago

Please clone this repo again. I added this function this morning after seeing this issue.

Thank you!