Where to find train_list.txt for blendedmvs

[JunyuanDeng]

Hello,

When I run train code, it appears:
dust3r_dataset/prepocessed_data/blendedmvs_processed/train_list.txt not found

How to create this file?

Thanks in advance!

[jedyang97]

Hi @JunyuanDeng , you can use the following script to generate the splits for BlendedMVS:

import os
import random

# Define the path to your dataset directory
base_dir = '/path/to/BlendedMVS'

# Fetch all scene folders
scene_dirs = [d for d in os.listdir(base_dir) if os.path.isdir(os.path.join(base_dir, d))]

# Set the random seed for reproducibility
random.seed(42)

# Shuffle the scene directories for random splitting
random.shuffle(scene_dirs)

# Perform a 90-10 split
split_index = int(0.9 * len(scene_dirs))
train_scenes = scene_dirs[:split_index]
test_scenes = scene_dirs[split_index:]

# Write the train list
with open(os.path.join(base_dir, 'train_list.txt'), 'w') as f:
    for scene in train_scenes:
        f.write(scene + '\n')

# Write the test list
with open(os.path.join(base_dir, 'test_list.txt'), 'w') as f:
    for scene in test_scenes:
        f.write(scene + '\n')

print(f"Train list saved with {len(train_scenes)} scenes.")
print(f"Test list saved with {len(test_scenes)} scenes.")

[jedyang97]

To make things easier, I attach the generated train_list.txt and test_list.txt here:

train_list.txt

test_list.txt