Part-Aware Data Augmentation for 3D Object Detection in Point Cloud

Jaeseok Choi

Last update: Jan 3, 2023

Related tags

Overview

Part-Aware Data Augmentation for 3D Object Detection in Point Cloud

This repository contains a reference implementation of our Part-Aware Data Augmentation for 3D Object Detection in Point Cloud.

If you find this code useful in your research, please consider citing our work:

@article{choi2020part,
  title={Part-Aware Data Augmentation for 3D Object Detection in Point Cloud},
  author={Choi, Jaeseok and Song, Yeji and Kwak, Nojun},
  journal={arXiv preprint arXiv:2007.13373},
  year={2020}
}

Prerequisites

Our code was tested on second.pytorch and OpenPCDet.
This repository contains only part-aware data augmentation code.
Refer to the link above for code such as data loader or detector.

Usage

Args:
    ** only supports KITTI format **
    points: lidar points (N, 4), 
    gt_boxes: ground truth boxes (B, 7),
    gt_names: ground truth classes (B, 1), 
    class_names: list of classes to augment (3),
    pa_aug_param: parameters for PA_AUG (string).

Returns:
    points: augmented lidar points (N', 4),
    gt_boxes_mask: mask for gt_boxes (B)


class_names = ['Car', 'Pedestrian', 'Cyclist']
pa_aug_param = "dropout_p02_swap_p02_mix_p02_sparse40_p01_noise10_p01"

pa_aug = PartAwareAugmentation(points, gt_boxes, gt_names, class_names=class_names)
points, gt_boxes_mask = pa_aug.augment(pa_aug_param=pa_aug_param)
gt_boxes = gt_boxes[gt_boxes_mask]

You might also like...

Style-based Point Generator with Adversarial Rendering for Point Cloud Completion (CVPR 2021)

Style-based Point Generator with Adversarial Rendering for Point Cloud Completion (CVPR 2021) An efficient PyTorch library for Point Cloud Completion.

119 Jan 2, 2023

Implementation of the "PSTNet: Point Spatio-Temporal Convolution on Point Cloud Sequences" paper.

PSTNet: Point Spatio-Temporal Convolution on Point Cloud Sequences Introduction Point cloud sequences are irregular and unordered in the spatial dimen

63 Dec 9, 2022

Implementation of the "Point 4D Transformer Networks for Spatio-Temporal Modeling in Point Cloud Videos" paper.

Point 4D Transformer Networks for Spatio-Temporal Modeling in Point Cloud Videos Introduction Point cloud videos exhibit irregularities and lack of or

101 Dec 29, 2022

Synthetic LiDAR sequential point cloud dataset with point-wise annotations

SynLiDAR dataset: Learning From Synthetic LiDAR Sequential Point Cloud This is official repository of the SynLiDAR dataset. For technical details, ple

78 Dec 27, 2022

[ICCV 2021 Oral] SnowflakeNet: Point Cloud Completion by Snowflake Point Deconvolution with Skip-Transformer

This repository contains the source code for the paper SnowflakeNet: Point Cloud Completion by Snowflake Point Deconvolution with Skip-Transformer (ICCV 2021 Oral). The project page is here.

65 Dec 26, 2022

Unofficial implementation of Point-Unet: A Context-Aware Point-Based Neural Network for Volumetric Segmentation

Point-Unet This is an unofficial implementation of the MICCAI 2021 paper Point-Unet: A Context-Aware Point-Based Neural Network for Volumetric Segment

9 Dec 7, 2022

Code for paper "ASAP-Net: Attention and Structure Aware Point Cloud Sequence Segmentation"

ASAP-Net This project implements ASAP-Net of paper ASAP-Net: Attention and Structure Aware Point Cloud Sequence Segmentation (BMVC2020). Overview We i

26 Aug 25, 2022

[ICCV 2021 Oral] PoinTr: Diverse Point Cloud Completion with Geometry-Aware Transformers

PoinTr: Diverse Point Cloud Completion with Geometry-Aware Transformers Created by Xumin Yu*, Yongming Rao*, Ziyi Wang, Zuyan Liu, Jiwen Lu, Jie Zhou

317 Dec 26, 2022

A modified version of DeepMind's Alphafold2 to divide CPU part (MSA and template searching) and GPU part (prediction model)

ParallelFold Author: Bozitao Zhong This is a modified version of DeepMind's Alphafold2 to divide CPU part (MSA and template searching) and GPU part (p

77 Dec 22, 2022

Comments

pvrcnn result in pa-aug paper lower than original paper

Hi @sky77764，thanks for sharing the code.

I have a question when reading the paper, moderate car val AP [email protected] of pvrcnn is 84.84 in original pvrcnn paper and 83.61 in open-pcdet implementation, while in pa-aug paper is 80.43.

Why is it much lower than original paper or open-pcdet implementation.

opened by turboxin 4

Inconsistent rotation directions in rotation_3d_in_axis and rotation_points_single_angle

In box_np_ops.py it's written as:

def rotation_points_single_angle(points, angle, axis=0):
    # points: [N, 3]
    rot_sin = np.sin(angle)
    rot_cos = np.cos(angle)
    if axis == 1:
        rot_mat_T = np.array(
            [[rot_cos, 0, -rot_sin], [0, 1, 0], [rot_sin, 0, rot_cos]],
            dtype=points.dtype)
    elif axis == 2 or axis == -1:
        rot_mat_T = np.array(
            [[rot_cos, -rot_sin, 0], [rot_sin, rot_cos, 0], [0, 0, 1]],
            dtype=points.dtype)
    elif axis == 0:
        rot_mat_T = np.array(
            [[1, 0, 0], [0, rot_cos, -rot_sin], [0, rot_sin, rot_cos]],
            dtype=points.dtype)
    else:
        raise ValueError("axis should in range")

    return points @ rot_mat_T

This results in mathematically negative rotation directions when axis == 2 or axis == 0, but mathematically negative direction for axis == 1. Transposing the matrices for axis == 2 and axis == 0 would fix that. The same happens in the rotation_3d_in_axis function.

For reference see Wikipedia on rotation matrices

opened by Tobi995 0

Part-Aware Data Augmentation for 3D Object Detection in Point Cloud

Related tags

Overview

Part-Aware Data Augmentation for 3D Object Detection in Point Cloud

Prerequisites

Usage

You might also like...

Style-based Point Generator with Adversarial Rendering for Point Cloud Completion (CVPR 2021)

Implementation of the "PSTNet: Point Spatio-Temporal Convolution on Point Cloud Sequences" paper.

Implementation of the "Point 4D Transformer Networks for Spatio-Temporal Modeling in Point Cloud Videos" paper.

Synthetic LiDAR sequential point cloud dataset with point-wise annotations

[ICCV 2021 Oral] SnowflakeNet: Point Cloud Completion by Snowflake Point Deconvolution with Skip-Transformer

Unofficial implementation of Point-Unet: A Context-Aware Point-Based Neural Network for Volumetric Segmentation

Code for paper "ASAP-Net: Attention and Structure Aware Point Cloud Sequence Segmentation"

[ICCV 2021 Oral] PoinTr: Diverse Point Cloud Completion with Geometry-Aware Transformers

A modified version of DeepMind's Alphafold2 to divide CPU part (MSA and template searching) and GPU part (prediction model)

Comments

pvrcnn result in pa-aug paper lower than original paper

Inconsistent rotation directions in rotation_3d_in_axis and rotation_points_single_angle

Owner

Jaeseok Choi

An integration of several popular automatic augmentation methods, including OHL (Online Hyper-Parameter Learning for Auto-Augmentation Strategy) and AWS (Improving Auto Augment via Augmentation Wise Weight Sharing) by Sensetime Research.

Scale-aware Automatic Augmentation for Object Detection (CVPR 2021)

Semantic Segmentation for Real Point Cloud Scenes via Bilateral Augmentation and Adaptive Fusion (CVPR 2021)

The source code for the Cutoff data augmentation approach proposed in this paper: "A Simple but Tough-to-Beat Data Augmentation Approach for Natural Language Understanding and Generation".

Image transformations designed for Scene Text Recognition (STR) data augmentation. Published at ICCV 2021 Workshop on Interactive Labeling and Data Augmentation for Vision.

Investigating Attention Mechanism in 3D Point Cloud Object Detection (arXiv 2021)

3D cascade RCNN for object detection on point cloud

Weakly Supervised 3D Object Detection from Point Cloud with Only Image Level Annotation

PointRCNN: 3D Object Proposal Generation and Detection from Point Cloud, CVPR 2019.

BADet: Boundary-Aware 3D Object Detection from Point Clouds (Pattern Recognition 2022)