GrooMeD-NMS: Grouped Mathematically Differentiable NMS for Monocular 3D Object Detection

GrooMeD-NMS: Grouped Mathematically Differentiable NMS for Monocular 3D Object Detection, CVPR 2021

Abhinav Kumar, Garrick Brazil, Xiaoming Liu

[project], [supp], [slides], [1min_talk], demo, arxiv

This code is based on Kinematic-3D, such that the setup/organization is very similar. A few of the implementations, such as classical NMS, are based on Caffe.

References

Please cite the following paper if you find this repository useful:

@inproceedings{kumar2021groomed,
  title={{GrooMeD-NMS}: Grouped Mathematically Differentiable NMS for Monocular {$3$D} Object Detection},
  author={Kumar, Abhinav and Brazil, Garrick and Liu, Xiaoming},
  booktitle={IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2021}
}

Setup

• Requirements

  1. Python 3.6
  2. Pytorch 0.4.1
  3. Torchvision 0.2.1
  4. Cuda 8.0
  5. Ubuntu 18.04/Debian 8.9

  This is tested with NVIDIA 1080 Ti GPU. Other platforms have not been tested. Unless otherwise stated, the below scripts and instructions assume the working directory is the project root.

  Clone the repo first:

  git clone https://github.com/abhi1kumar/groomed_nms.git

• Cuda & Python

  Install some basic packages:

  sudo apt-get install libopenblas-dev libboost-dev libboost-all-dev git
  sudo apt install gfortran
  
  # We need to compile with older version of gcc and g++
  sudo apt install gcc-5 g++-5
  sudo ln -f /usr/bin/gcc-5 /usr/local/cuda-8.0/bin/gcc
  sudo ln -s /usr/bin/g++-5 /usr/local/cuda-8.0/bin/g++

  Next, install conda and then install the required packages:

  wget https://repo.anaconda.com/archive/Anaconda3-2020.02-Linux-x86_64.sh
  bash Anaconda3-2020.02-Linux-x86_64.sh
  source ~/.bashrc
  conda list
  conda create --name py36 --file dependencies/conda.txt
  conda activate py36

• KITTI Data

  Download the following images of the full KITTI 3D Object detection dataset:

  • left color images of object data set (12 GB)
  • camera calibration matrices of object data set (16 MB)
  • training labels of object data set (5 MB)

  Then place a soft-link (or the actual data) in data/kitti:

   ln -s /path/to/kitti data/kitti

  The directory structure should look like this:

  ./groomed_nms
  |--- cuda_env
  |--- data
  |      |---kitti
  |            |---training
  |            |        |---calib
  |            |        |---image_2
  |            |        |---label_2
  |            |
  |            |---testing
  |                     |---calib
  |                     |---image_2
  |
  |--- dependencies
  |--- lib
  |--- models
  |--- scripts

  Then, use the following scripts to extract the data splits, which use soft-links to the above directory for efficient storage:

  python data/kitti_split1/setup_split.py
  python data/kitti_split2/setup_split.py

  Next, build the KITTI devkit eval:

   sh data/kitti_split1/devkit/cpp/build.sh

• Classical NMS

  Lastly, build the classical NMS modules:

  cd lib/nms
  make
  cd ../..

Training

Training is carried out in two stages - a warmup and a full. Review the configurations in scripts/config for details.

chmod +x scripts_training.sh
./scripts_training.sh

If your training is accidentally stopped, you can resume at a checkpoint based on the snapshot with the restore flag. For example, to resume training starting at iteration 10k, use the following command:

source dependencies/cuda_8.0_env
CUDA_VISIBLE_DEVICES=0 python -u scripts/train_rpn_3d.py --config=groumd_nms --restore=10000

Testing

We provide logs/models/predictions for the main experiments on KITTI Val 1/Val 2/Test data splits available to download here.

Make an output folder in the project directory:

mkdir output

Place different models in the output folder as follows:

./groomed_nms
|--- output
|      |---groumd_nms
|      |
|      |---groumd_nms_split2
|      |
|      |---groumd_nms_full_train_2
|
| ...

To test, run the file as below:

chmod +x scripts_evaluation.sh
./scripts_evaluation.sh

Contact

For questions, feel free to post here or drop an email to this address- [email protected]