But, I still wonder want format my processed dataset should look like.

[Ye110ws4ar]

Generally, I chose the images of 00~03 cameras with each contains 3 frames, the .ply file, and the transforms.json which stores the para of 8 cameras out of the given 0023_06 dataset . Easily, the skeleton files of 8x3 images were created, so were 4x3 generated fmasks. What I want to achieve is to get the other 4x3 images as output with the known skeleton files and 4x3 images as input, so I edited the demo_4d_tiny.yaml as below:
sampler: spa_label_range: [0, 8, 1] # target cameras tem_label_range: [0, 4, 1] # target frames input_spa_labels: [0,1,2,3] # input cameras window_size : 2
Sadly, the error really makes me confused. It suggests that I should also put the wanted 4x3 images and fmasks as input to let the inference.py run correctly :(
Should I have got the wanted images, what's the meaning of the whole processing? So, how could I set the parameter correctly and rearrange my input format?
My input format now:
-fmasks
-00
-000000.png
-000001.png
-000002.png
-01
-02
-03
-images
-00
-000000.webp
-000001.webp
-000002.webp
-01
-02
-03
-skeletons (generated of transforms.json which stores 8 cameras)
-00
-000000.png
-000001.png
-000002.png
-01
-02
-03
-04
-05
-06
-07
-transform.json