Inside data_prep please use this tree for running in custom dataset (within some days i will write code so that we can use seperate datasets/.py files for data loader part in custom dataset training)
Data|
|
|
|--> ScanNet (clone from scanet repo)
|
|--> scans|-->scene0003_02|-->annotation----------->Segmentation----->0 to 1000 images with w,h,3 shape
| | |---> plane_info.npy (usage is optional)
| | |---> planes.npy(use plan parameters)
| | |
| | |
| | |
| | |-->frames--------------->color 0 to 1000 images with w,h,3 shape
| | | --------------->depth 0 to 1000 images with w,h,3 shape
| | | --------------->intrinsic (default)
| | | ----------->pose (default) or else could be got from using "--testingIndex=0" for each image | | |
| | |
| | |
| | |-->scene0003_02.txt
|
|--> tasks|
| |----> Scannetv2-labels.combined.tsv
|
|--> anchor_planes_N.npy
|
|-->invalid_indices_test.txt
|
|--> invalid_indices_train.txt
|
|