• This code was developed with Python 3.7.10 & Pytorch 1.8.1 & CUDA 11.3
• Other requirements: numpy, cv2, tensorboardX
• Clone this repo

git clone https://github.com/ccc870206/3D-PL.git
cd 3D-PL

Target dataset: KITTI

Rename the main folder of kitti dataset as kitti_data and put the folder under data/

data
  |----kitti_data 
         |----2011_09_26         
         |----2011_09_28        
         |----......... 

Source dataset: vKITTI (1.3.1)

Download our pre-trained model and put the folder under checkpoints/.

• Test the model pre-trained with single-image setting

python3 test.py --model test --name best_model_single_image --which_epoch best

• Test the model pre-trained with stereo-pair setting

python3 test.py --model test --name best_model_stereo_pair --which_epoch best

Code is inspired by T^2Net and GASDA.