Appendix B: Datasets and Test Cases

This section introduces the dataset formats supported by the project and how to use test cases.

Dataset requirements

Default support: KITTI dataset

• KITTI is a standard dataset in the field of autonomous driving, containing rich binocular image data.

Data organization format

The project requires that the data be organized in the following structure:

data/
├── left/
│ ├── 0000000000.png
│ ├── 0000000001.png
│ └── ...
└── right/
├── 0000000000.png
├── 0000000001.png
└── ...

Data description

1. **Left image (left): Left camera image used for disparity matching.

2. **Right image (right): Right camera image for disparity matching.

Test case

Test script provided

You can run the test through the script in the scripts/ folder:

python test_stereo.py

Test results

After running the program, it will display:

1. Disparity map

2. Stixel visualization

3. Bounding box annotation map

Custom data

When using a custom dataset, please:

1. Organize the image data in the above format.

2. Adjust the reading path in main.cpp:

   sprintf(chLeftImageName, "./data/left/%010d.png", cntFrame);
   sprintf(chRightImageName, "./data/right/%010d.png", cntFrame);