Glass Detection for Depth Completion Lite ver.
Publication: Glass Area Detection and Linear Interpolation for Indoor Depth Completion, 2023. Link
- Datasets Source: https://github.com/tsunghan-wu/Depth-Completion
--> Data_140 include GT Data and Render Data, total 140 pics.
--> no_glass_275 is non-glass dataset.
--> txt file include raw depth file paths, separate to GT Data and Render Data, modify it as needed.
-
Preprocessing: matterport3D's data include noise, require to execute clean_depth_noise.py first.
-
find glass area(output: uint8): glasscut.py
-
depth completeion method: glass_depth_completion.py
--> input file path for raw depth, reprocessing depth, mask.
--> input file path for saving complete depth files. -
glass area metrics: eval_glass_mask_metrics.py
-
depth completion metrics: eval_depth_completion_metrics.py
-
Ablation study in different methods:
OpenCV inpainting: cv_inpainting.py
Scipy griddata: griddata.py -
npy file to exr file: BDepth_npy_to_exr.py
-
exr file to png file: convert_exr_depth_to_png.py
-
gray png file to rgb color: show_rgb_depth.py