Skip to content

The purpose of this research is to do 3D object detection from a photo using machine learning. The goal is a working model that can detect multiple 3D objects and provide their dimensions and orientation given a photo. There are many ways to implement this, but this project uses center-based 3D object detection and tracking, where the model pred…

Notifications You must be signed in to change notification settings

umd-fire-coml/2020-3D-Object-Detection

Repository files navigation

2020-3D-Object-Detection

Abstract

3D Object Detection with CenterPoint The purpose of this research is to do 3D object detection from a photo using machine learning. The goal is a working model that can detect multiple 3D objects and provide their dimensions and orientation given a photo. There are many ways to implement this, but this project uses center-based 3D object detection and tracking, where the model predicts the physical, three-dimensional center of detected objects, which is regressed to predict the dimensions and orientation of the object. This output can be tracked over time to predict the velocity of objects and visualize their movement using a GIF. The model was found to have a high level of accuracy in detecting and tracking 3D objects. This model can be used to detect and track objects in a plethora of applications including surveillance, autonomous vehicles, and augmented reality.

Short video about our project

Watch the video

File descriptions

checker

  • data-checker--file-output-MASTER.txt: has list of data files used in validation
  • data-checker--file-output.py: script used to create MASTER file
  • validation.sh: script used to run validation

data

  • empty folder where the data will be put

environment

  • requirements.txt: list of all the packages needed to download

processor

  • download.sh: script for downloading the data

configs/centerpoint

  • has 3 different models that you can pick from for training

tools

  • This folder contains utilities for the demo/testing like visualizing and turning pictures into a gif

det3d

  • This folder is the framework used in our project

work_dirs/centerpoint_pillar_512_demo

  • last.pth: pretrained weights used for the demo and can be used for testing

Training

For instructions on how to get training started go to this google colab

Citations

@article{Yin2020Centerbased3O, title={Center-based 3D Object Detection and Tracking}, author={Tianwei Yin and Xingyi Zhou and Philipp Kr{"a}henb{"u}hl}, journal={ArXiv}, year={2020}, volume={abs/2006.11275} }

@article{zhou2018voxelnet, title={VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection}, journal={CVPR}, author={Zhou, Yin and Tuzel, Oncel}, year={2018}, }

About

The purpose of this research is to do 3D object detection from a photo using machine learning. The goal is a working model that can detect multiple 3D objects and provide their dimensions and orientation given a photo. There are many ways to implement this, but this project uses center-based 3D object detection and tracking, where the model pred…

Resources

Stars

Watchers

Forks

Packages

No packages published