Official Code of LFCNet

Installation

we run the project on PyTorch==2.0.0 on CUDA 11.8 and Python==3.10.12

Step 1. Clone and install requirements

git clone https://github.com/rangganast/LFCNet.git
cd LFCNet
pip install -r requirements.txt

Step 2. Install Pytorch-Correlation-extension

cd ..
git clone https://github.com/ClementPinard/Pytorch-Correlation-extension.git
cd Pytorch-Correlation-extension
python3 setup.py install

Step 3. Download the weights

• Err: 1.5 | 20
• Err: 1.0 | 10
• Err: 0.5 | 5
• Err: 0.2 | 2

Data Preparation

The data folders are organized as follows:

├── KITTI-360/
|   └── calibration
|   └── data_2d_raw
|        └── 2013_05_28_drive_0000_sync  
|            └── data
|               └── 000000.png
|               └── 000001.png
|               └── 000002.png
|               └── ...
|   └── data_3d_raw
|       └── 2013_05_28_drive_0000_sync  
|           └── data
|               └── 000000.bin
|               └── 000001.bin
|               └── 000002.bin
|               └── ...

Testing

Put the weights in checkpoint_weights/run, and run:

python test_iterative.py

For iterative refinement evaluation, and

python test_continuous.py

For multi frame analysis

Contact

For questions about our paper or code, please contact rangganast@gmail.com.