maybeLx / MVSFormerPlusPlus

Codes of MVSFormer++: Revealing the Devil in Transformer’s Details for Multi-View Stereo (ICLR2024)
Apache License 2.0
178 stars 6 forks source link

Test on my own data #29

Open xuncpp opened 3 weeks ago

xuncpp commented 3 weeks ago

First of all, thank you for your excellent work and detailed instructions. I defined my own dataset according to your steps and ran the code but got an error. The error should be because my dataset does not have mask_filename_hr and depth_filename_hr.

if i == 0: # reference view
if self.kwargs['dataset'] == 'dtu':
mask_read_ms = self.read_mask_hr(mask_filename_hr)
depth_ms = self.read_depth_hr(depth_filename_hr)
mask = mask_read_ms
depth_values ​​= np.arange(depth_min, depth_interval * (self.ndepths - 0.5) + depth_min, depth_interval, dtype=np.float32)

How could I solve this problem? Thanks again!

maybeLx commented 2 weeks ago

Apologies for the bug. I've fixed the issue in the Test on your own data section. Since our model is trained on the DTU dataset, we require depth and mask files during training. However, these files are not needed for testing. You can set the dataset parameter to alternatives, such as My_dataset.