mvmm

Implementation of MVMM in PyTorch for KITTI 3D Object Detetcion

Acknowledgement

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

Installation

• Clone this repository

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

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

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

• Install spconv

  pip install spconv-cu102
  

• Compile external modules

  cd mvmm
  python setup.py develop
  

KITTI3D Dataset (41.5GB)

• Download KITTI3D dataset: calib, velodyne, label_2 and image_2.
• Download road plane for data augmentation.
• Organize the downloaded files as follows

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

• Generate ground truth databases and data infos by running the following command

  # This will create two database dirs and six info files in mvmm/data/kitti (Take 20 mins).
  cd mvmm
  python -m data.kitti_dataset create_kitti_infos
  

• Display the dataset

  # Display the dataset and show annotations in the range image
  python dataset_player.py --augment_data --show_boxes --onto_range_image
  
  # Display the dataset and show annotations in point clouds
  python dataset_player.py --augment_data --show_boxes
  

Demo

• Run the demo with a trained model

  # Show detections in the RGB image
  python demo.py --checkpoint=checkpoints/checkpoint_epoch_80.pth --show_boxes --onto_image
  
  # Show detections in point clouds
  python demo.py --checkpoint=checkpoints/checkpoint_epoch_80.pth --show_boxes
  

Training

• Train your model using the following commands

  python train.py
  

Evaluation

• Evaluate your model using the following commands

  python test.py