├── img
    ├── intro.pdf
    └── intro.png
└── README.md


/img/intro.pdf:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/BFZD233/TranScene/HEAD/img/intro.pdf


--------------------------------------------------------------------------------
/img/intro.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/BFZD233/TranScene/HEAD/img/intro.png


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
 1 | # Multi-label Stereo Matching for Transparent Scene Depth Estimation
 2 | 
 3 | 
 4 | ![intro](./img/intro.png "intro")
 5 | ## News
 6 | 
 7 | - **2025.5.21**: The code will be uploaded in a few weeks.
 8 | 
 9 | 
10 | ## Installation
11 | 
12 | Our code is based on CUDA 11.7 and PyTorch 2.0.1. We recommend using Anaconda to create a new environment:
13 | 
14 | ```bash
15 | conda create -n trans python=3.7
16 | conda activate trans
17 | ```
18 | 
19 | Then, install the dependencies:
20 | 
21 | ```bash
22 | pip install -r requirements.txt
23 | ```
24 | 
25 | ## Dataset
26 | 
27 | We use the [TartanAir](https://github.com/castacks/tartanair_tools), [SceneFlow](https://lmb.informatik.uni-freiburg.de/resources/datasets/SceneFlowDatasets.en.html), and [Raw data of KITTI](https://github.com/youmi-zym/TemporalStereo?tab=readme-ov-file#kitti-20122015) datasets in our experiments. Please download the datasets and organize them as follows:
28 | 
29 | ```
30 | ├── datasets
31 |     ├── sceneflow
32 |         ├── driving                                               
33 |         │   ├── disparity                                         
34 |         │   ├── frames_cleanpass                                  
35 |         │   └── frames_finalpass                                  
36 |         ├── flying3d                                              
37 |         │   ├── disparity                                         
38 |         │   ├── frames_cleanpass                                  
39 |         │   └── frames_finalpass                                  
40 |         └── monkaa                                                
41 |             ├── disparity                                         
42 |             ├── frames_cleanpass                                             
43 |             └── frames_finalpass
44 |     ├── TranScene
45 |         ├── left
46 |         │   ├── disrparity
47 |         │   ├── disparity_without_trans 
48 |         │   └── img
49 |         └── right
50 | ```
51 | 
52 | ## Checkpoints
53 | 
54 | To Do.
55 | 
56 | ## Evaluation
57 | 
58 | Before evaluation, please download the checkpoints and put them in the `./checkpoints` directory.
59 | 
60 | You can evaluate the pre-trained models on TranScene by running the following scripts:
61 | 
62 | ```bash
63 | bash trans_evaluate.sh
64 | ```
65 | 
66 | ## Training
67 | 
68 | For SceneFlow dataset, you can train the model by running the following script:
69 | ```bash
70 | bash sceneflow_train.sh
71 | ```
72 | 
73 | For TranScene dataset, you can train the model by running the following script:
74 | ```bash
75 | bash trans_train.sh
76 | ```
77 | 
78 | ## Acknowledgement
79 | 
80 | Our code is based on [RAFT-Stereo](https://github.com/princeton-vl/RAFT-Stereo). We thank the authors for this great work.


--------------------------------------------------------------------------------