prepare the ground truth depth maps

[wwj-53]

Based on your description，the directory of dataset structure is shown as ：
/path/to/endovis_data/
dataset1/
keyframe1/
image_02/
data/
0000000001.png

but after processing according to your method, it looks like this:
/path/to/endovis_data/
dataset1/
keyframe1/
data/
0000000001.png
There is no image folder name available.

When I run CUDA_VISIBLE_DEVICES=0 python export_gt_depth.py --data_path /media/wwj/aj_new/dataset_2 --split endovis，
The error is as follows：
(endodac) wwj@cpnt:~/3DGS/AF-SfMLearner$ CUDA_VISIBLE_DEVICES=0 python export_gt_depth.py --data_path /media/wwj/aj_new/dataset_2 --split endovis
Exporting ground truth depths for endovis
1
dataset3/keyframe4
[ WARN:0@0.005] global loadsave.cpp:241 findDecoder imread_('/media/wwj/aj_new/dataset_2/dataset3/keyframe4/data/groundtruth/scene_points000389.tiff'): can't open/read file: check file path/integrity
Traceback (most recent call last):
File "/home/wwj/3DGS/AF-SfMLearner/export_gt_depth.py", line 61, in
export_gt_depths_SCARED()
File "/home/wwj/3DGS/AF-SfMLearner/export_gt_depth.py", line 48, in export_gt_depths_SCARED
depth_gt = depth_gt[:, :, 0]
TypeError: 'NoneType' object is not subscriptable

I have changed line 44 from "image-02/data/roundtruth" to "data/roundtruth" (otherwise it will not run according to the structure of my dataset)

[suxuanya]

I also encountered this problem. Have you solved it?

[wwj-53]

I also encountered this problem. Have you solved it?

no

[ShuweiShao]

Hi all,

Please create the image_02 folder by yourself and put the data in it. The format follows the KITTI dataset and ''image_02'' indicates the left frames.

[suxuanya]

Thanks a lot.