search

chensjtu / GaussianObject

GaussianObject: High-Quality 3D Object Reconstruction from Four Views with Gaussian Splatting (SIGGRAPH Asia 2024, TOG)
914 stars 56 forks source link

Rabbit dataset #52

Closed seulqxq closed 1 month ago

seulqxq commented 1 month ago

Hi, I want to train using the rabbit dataset, using the mask generated by segment_anything.ipynb, but running python pred_poses.py -s data/realcap/rabbit --sparse_num 4 gets wrong. image Here is the data: image

TXSevenXT commented 1 month ago

I've got this problem :( For the moment, no solution :(

TXSevenXT commented 1 month ago

Find a solution in deleting "0" in the code :
original_masks = [np.array(Image.open(mask).resize(image.size))[:, :, 0] / 255.0 for mask, image in zip(masks, original_images)] Becomes original_masks = [np.array(Image.open(mask).resize(image.size))[:, :,] / 255.0 for mask, image in zip(masks, original_images)]

I will try with my others pictures and tell you if it's ok ^^