ldfmm

Implementation of LDFMM in PyTorch for KITTI 3D Object Detetcion

Acknowledgement

• This repository is developed based on xinzhuma and open-mmlab's work.

Installation

• Clone this repository

  git clone git@github.com:shangjie-li/ldfmm.git
  

• Install PyTorch environment with Anaconda (Tested on Ubuntu 16.04 & CUDA 10.2)

  conda create -n ldfmm python=3.6
  conda activate ldfmm
  cd ldfmm
  pip install -r requirements.txt
  

• Compile external modules

  cd ldfmm
  python setup.py develop
  

KITTI3D Dataset (41.5GB)

• Download KITTI3D dataset: calib, velodyne, label_2 and image_2.
• Organize the downloaded files as follows

  ldfmm
  ├── data
  │   ├── kitti
  │   │   │── ImageSets
  │   │   │   ├──test.txt & train.txt & trainval.txt & val.txt
  │   │   │── training
  │   │   │   ├──calib & velodyne & label_2 & image_2
  │   │   │── testing
  │   │   │   ├──calib & velodyne & image_2
  ├── helpers
  ├── layers
  ├── ops
  ├── utils
  

• Display the dataset

  # Display the dataset and show annotations in the image
  python dataset_player.py --augment_data --show_keypoints --show_boxes2d --show_boxes3d
  
  # Display the dataset and show annotations in point clouds
  python dataset_player.py --augment_data --show_boxes3d --show_lidar_points
  
  # Display the dataset and show the encoded heatmap
  python dataset_player.py --augment_data --show_heatmap
  

Demo

• Run the demo with a trained model

  # Show detections in the image
  python demo.py --checkpoint=checkpoints/checkpoint_epoch_80.pth --show_boxes3d
  
  # Show detections in point clouds
  python demo.py --checkpoint=checkpoints/checkpoint_epoch_80.pth --show_boxes3d --show_lidar_points
  

Training

• Train your model using the following commands

  python train.py
  

Evaluation

• Evaluate your model using the following commands

  python test.py