If you find the poject helps you, you can cite our paper:
Cao D, Yue H, Liu Z, et al. BEVLCD+: Real-Time and Rotation-Invariant Loop Closure Detection Based on BEV of Point Cloud[J]. IEEE Transactions on Instrumentation and Measurement, 2023.
Yue H, Cao D, Liu Z, et al. Cross Fusion of Point Cloud and Learned Image for Loop Closure Detection[J]. IEEE Robotics and Automation Letters, 2024.
We provide code for BEV mode and fusion mode, so you can easily train and test.
Before you can use this project, you'll need to do the following:
Prepare Dataset Structure: Use preparedataset.py to construct a dataset structure that complies with the project's requirements. Make sure to update the necessary paths in the code.
Prepare environment: Use the commonds on env.txt to create your environment. Windows and Ubuntu is OK.
To run the code, follow these steps:
Configure the code to run in either BEV mode or fusion mode using the settings in config.yaml.
If you want to load a trained model used in the paper, ensure that you update the file path accordingly.
Run python train.py
Evaluate the saved data using the evaluation script.
If you have any questions please feel free to contact us.